Escolar Documentos
Profissional Documentos
Cultura Documentos
INTRODUCTION ................................................................................................................... 3
I. UNRECOGNIZED RISK AND BOND RATINGS ................................................................. 3
II. STRUCTURAL CHANGES IN THE BOND RATING INDUSTRY: THE MOVE FROM PASSIVE
TO ACTIVE ................................................................................................................... 7
A. A Brief Background on the Role of Credit Rating Agencies ................................ 7
B. Are Credit Rating Agencies Still Publishers or Are They Underwriters?.......... 11
C. New ConflictsThe Buy Side ............................................................................ 17
D. The Problems With Ratings In Structured FinanceEvolving Models ............. 19
E. Regulatory Issues............................................................................................... 28
F. Rating Agencies Are Activist in Ways They Have Never Been........................... 31
G. Other Legal Risk Securitization Law ................................................................. 33
III. TRADITIONAL BOND RATINGS DO NOT PROPERLY ACCOUNT FOR RESIDENTIAL
MORTGAGE-BACKED SECURITY RISKS ...................................................................... 34
A. Pooling Mortgages does not create Diversification .......................................... 35
B. Current Bond Ratings Methods are Misleading when Applied to RMBS .......... 36
1. Corporate Default Risk is a Function of Investment Decisions, while
RMBS Default Risk is a Function of Investment Performance ................... 37
2. Performance Point Estimates that are the Basis for Corporate Bond
Ratings are not Relevant for RMBS ............................................................ 38
3. Cumulative Losses Can Only Increase in Structured Finance Debt.......... 40
4. The Statistical Distribution of Mortgage Pool Performance Is Skewed
Relative to that of Corporate Investments................................................... 41
5. Mortgage Pool Cumulative Loss Distributions Narrow over Time ............ 43
6. Implications: When Mortgage Pools Do Not Perform, the RMBS Will
Not Be Promptly Downgraded by Traditional Ratings Methods ................ 44
INTRODUCTION
Conventional wisdom holds that market volatility or market risk has declined.
For risk that remains, markets have reduced the price for taking on that risk.
Figure 1 illustrates the conventional view as it related to mortgage markets. Since
2003, swaption volatility has decreased substantially, along with the Mortgage
Bankers Association Refinancing cost index and residential mortgage-backed
security (RMBS) yield spreads over LIBOR. Such circumstances have reduced
mortgage costs to homebuyers and set the stage for the new financial markets,
that are thought to more efficiently value and price risk. Of course, risk does not
4 Joshua Rosner & Joseph R. Mason
Source: Nomura, Structured Finance Trends Yield Spreads, Credit Support, and Collateral
Performance The Big Picture, June 27, 1005, at 9.
Source: Fitch IBCA, Rating U.S. Residential Subprime Mortgage Securities, Jul. 18, 2001.
May 2007 Mortgage-Backed Security Ratings 5
the increased gradation of risk among mortgage borrowers. The additional grades
of risk arise from the willingness to underwrite mortgages for more risky
borrowers, encouraged by the democratization of credit since the 1970s, as well
as making riskier loans to more typical borrowers. Both practices increase the
total amount of risk to be sold in the marketplace. However, increased grading of
risk induced increased complexity, and therefore increased opacity. Risk that is
more difficult to see, by virtue of complexity, is risk just the same.
There are plenty of reasons to believe that the amount of risk in the
marketplace has increased while opacity has made it seem otherwise. Figure 3
shows that defaults on ABS and RMBS increased substantially between 1991 and
2003 and declined sharply thereafter. During that decline, however, the ratings
Source: FitchIBCA, Fitch Global Structured Finance 1991-2005 Default Study, Nov. 26, 2006.
agencies have continued to revise their loss expectations to account for the
dynamics of the mortgage meltdown. For instance, on March 27, Standard &
Poors raised its expectation for losses on 2006 subprime mortgage bond issues
to as high as 7.75 percent from a previous peak assumption of about 6.5 percent.1
Moodys made similar revisions on April 20, 2007. As a result, both Moodys
and Standard & Poors parent corporation stock prices are down over the past
three months, partially due to reputational concerns.2
Of course, it is up to the reader to translate the revised loss expectations into
default scenarios. Figure 4 provides that translation from Fitchs Residential
Mortgage-Backed Securities Criteria. Ratings agency criteria are the
documentation circulated by ratings agencies to describe their loss analysis and
ratings methods. Figure 4 shows that Fitch anchors their expectations of the
relationship between losses and ratings changes in a number of academic studies
as well as their own research. The data are taken from criteria published well
before the meltdown, offering insights untainted by recent adjustments to
ratings agency methods meant to better capture risk (next time). Figure 4
illustrates that loss levels of 7.7 percent are expected to begin to impose losses on
investment-grade securities.
1. S&P raises 2006 subprime mortgage loss expectation, REUTERS, Mar. 27, 2007.
2. Serena Ng, Subprime Cloud Overshadows S&P, Moodys, WALL ST. J., Apr. 24,
2007, at C1.
6 Joshua Rosner & Joseph R. Mason
Data from the Mortgage Bankers Association suggests that losses may
continue to mount. The MBA Survey shows that many states are already
illustrating foreclosure and near foreclosure rates levels not seen since the thrift
crisis. Mortgage Bankers Association data from the third quarter of 2006 showed
that Ohio had the highest foreclosure rate in the country, while other Midwestern
states with fairly low home price appreciation over the last decadenamely,
Michigan, Indiana, Iowa, and Kentuckywere also among the highest. At that
time, California was below the U.S. average. States like Arizona and Nevada
ranked among the lowest, but foreclosures in those states are accelerating.
What is interesting about these trends is that while as early as January 2005,
delinquency and foreclosure data pointed to substantial deterioration in the
overall credit performance of even fixed-rate prime mortgage loans, actual losses
on securitized pools were extremely low during that period. Hence, S&P reported
981 RMBS upgrades and only RMBS 17 downgrades in 2004, and Moodys
reported 414 RMBS upgrades to only 4 RMBS downgrades for the same period.
That is the kind of news that led many to believe in decreased market risk. Figure
5 suggests the trend toward increased home price appreciation could be at its end,
leaving losses to accumulate in pools for some time.
It bears emphasis that the strong home-price appreciation since 1998 is the
principal factor that prevented delinquencies and defaults from developing into
losses before 2007.3 Figure 6 shows that markets are therefore beginning to price
risk in the housing sector with strong increases in spreads on almost all types of
securities.
Source: Nomura, Securitization & Real Estate Update, Feb. 28, 2007.
The following sections discuss elements of risk that were missed by investors
and ratings agencies and what can be done to better evaluate risk going forward
from the present meltdown. We explain that movements toward transparency for
these fundamental factors can go a long way to ensuring stability for the socially
and economically important segment of the U.S. economy, one that is crucial to
both consumer well-being and economic growth.
II. STRUCTURAL CHANGES IN THE BOND RATING INDUSTRY: THE MOVE FROM
PASSIVE TO ACTIVE
Since the creation of the modern credit rating industry by John Moody in
1909, ratings were offered to investors to assess credit-risks to the ability of a
dynamic institution to meet its financial obligations. Historically, the rating
agencies revenues were generated by subscription fees from subscribers who
received research and ratings on the creditworthiness of issuers of debt
securities.4
The value of ratings to investors is generally assumed to be a benchmark of
comparability it offers investors in differentiating between securities. Credit
rating agencies (CRAs) have long argued that the ratings scales they employed
were consistent across assets and markets. Not long ago Moodys stated The
need for a unified rating system is also reflected in the growing importance of
10. Bank for International Settlements, Committee on the Global Financial System,
The Role of Ratings in Structured Finance: Issues and Implications 24 (Jan. 2005)
(Model risk is not confined to structured finance. However, given the lack of
historical default data and the analytical challenges in assessing credit risk exposures (e.g.
10 Joshua Rosner & Joseph R. Mason
generally a positive movement for lenders, but as competition for loans increased
the profitability of traditional lending decreased.15 Hence, securitization became
a driver for new mortgage product development, which helped increase
homeownership rates and home prices in the past decade.
The changes in the role of rating agencies, as their business has evolved from
the rating of dynamic and on-going enterprises to static and defined-lived
structured assets, warrants inquiry into not only the rigors of their structured
finance rating models but also their historic claims of being publishers protected
by the First Amendments protections of the freedom of the press.16 Historically
rating agencies have claimed that their letter-grade ratings were merely an
opinion of the creditworthiness of an issuer or, according to a Fitch general
counsel, the worlds shortest editorial.17 Courts have found that ratings are
speech and, absent special circumstances, are protected by the First
Amendment [and] as a matter of public concern, would receive the heightened
protection of the actual malice standard.18 Given the changing nature of their
business and compensation there may be reason to question whether, today, they
should be viewed as editorializing or advertising.
The Senate Commerce Committee offered the following opinion in its Enron
report: The credit rating agencies seem to be trying to walk a fine line between
maintaining enormous market power through both official and unofficial uses of
their ratings, and insisting that their ratings are purely their opinion, and
therefore pure speech under a First Amendment analysis.19 Although the authors
do not claim legal expertise, there are clear reasons to consider whether, in
structured finance, the commercial speech of the CRAs would be protected upon
a challenge.
Rating agency claims of being publishers is called into question, by
implication, in a recent report to the European Commission of Securities
Regulators The CESR highlighted several areas where the rating agencies were
either not in compliance with the code or were questionably in compliance
16. Id.
16. See, e.g., Exposure to Litigation Related to Moodys Rating Opinions,
http://sec.edgar-online.com/2007/03/01/0001193125-07-043002/Section2.asp (last visited
Apr. 30, 2007) (Moodys faces litigation from time to time from parties claiming
damages relating to ratings actions. In addition, as Moody's international business
expands, these types of claims may increase because foreign jurisdictions may not have
legal protections or liability standards comparable to those in the U.S. (such as
protections for the expression of credit opinions as is provided by the First Amendment).
These risks often may be difficult to assess or quantify and their existence and magnitude
often remains unknown for substantial periods of time.).
17. Financial Oversight of Enron: The SEC and Private-Sector Watchdogs Report of
the Staff to the Senate Committee on Governmental Affairs Oct. 8, 2002 [hereinafter
Financial Oversight of Enron].
18. See, e.g., County of Orange v. McGraw Hill Cos., Inc., 245 B.R. 151, 156 n.4
(C.D. Cal. 1999).
19. Financial Oversight of Enron, supra note 17, at 123.
12 Joshua Rosner & Joseph R. Mason
with the code.20 Although the rating agencies seemed to be compliant with Code
1.1 under Quality and integrity of the rating process, the manner in which they
complied appears not to meet journalistic standards of ethics.21 The IOSCO Code
explains that the CRA should adopt, implement and enforce written procedures
to ensure that the opinions it disseminates are based on a thorough analysis of all
information known to the CRA that is relevant to its analysis according to the
CRAs published rating methodology.22 In its submissions, Fitch submitted that
Fitch shall have no obligation to verify or audit any information provided to it
from any source or to conduct any investigation or review, or to take any other
action, to obtain any information that the issuer has not otherwise provided to
Fitch.23 Although Moodys submission in answer to their relevant policy was
not direct, it appears that they also fall short of journalistic standards.24
Beyond the issues raised specifically by the CESR staff, there may be reason to
question the spirit with which one or more agencies comply with other provisions
of the code.25 It is not clear whether an agencys decision to recuse itself of the
responsibility to verify information provided by an issuer fully meets the implied
standard. Nor is it clear whether such conduct falls short of the standards the
agencies are expected to meet as investment advisors under the 1940 Act?26
The need for rating agencies to objectively assess and verify information rises in
structured finance transactions since, unlike the traditional ratings process in
which an enterprise can do little to change its risk characteristics in anticipation
of an issuance, in structured finance, the rating agency is an active part of the
structuring of the deal.27 In practice, arrangers will routinely use the rating
agencies publicly available models to pre-structure deals and subsequently
engage in a process that is iterative and interactive,28 informing the issuer of
the requirements to attain desired ratings in different tranches and largely
defining the requirements of the structures to achieve target ratings.293031
agency indicating the factors that need to be addressed to obtain the desired rating. In
particular, the agency has an indirect influence on how the tranches are configured to
ensure that the senior issue obtains the highest possible rating.) [hereinafter Authorite
des Marches Financiers Report]
32. Role of Ratings, supra note 10, at 38.
33. John S. Dzienkowski & Robert J. Peroni, The Decline in Lawyer Independence:
Lawyer Equity Investments in Clients 80 n.353 (2002), available at
http://www.law.fsu.edu/faculty/2001-2002workshops/dzienkowski.pdf.
34. Securities Act of 1933 2(a)(11), available at
http://sec.gov/divisions/corpfin/33act/sect2.htm.
35. Naomi O. Harden et al. v. Raffensperger, Hughes & Co., Inc., 94 U.S. 2892 (7th
Cir. 1995), available at http://caselaw.lp.findlaw.com/cgi-
bin/getcase.pl?court=7th&navby=case&no=942892.
36. Harden v Raffensperger, Hughes & Co., Inc., available at
http://caselaw.lp.findlaw.com/scripts/getcase.pl?court=7th&navby=case&no=942892.
37. Memorandum from Cahill Gordon & Reindel LLP on the constitutionality of
H.R. 2990, the Credit Rating Agency Duopoly Relief Act of 2005 8 (July 2005),
available at http://www2.standardandpoors.com/spf/pdf/media/Exhibit_2.pdf.
38. 17 C.F.R. 230.436.
May 2007 Mortgage-Backed Security Ratings 15
securities of this fact.39 It is, however, unclear whether the SECs exemption is
within the boundaries of Congressional intent given the fundamental changes in
the use of ratings since they were given such protection.40
Moreover, if a rating agencys role in an issuance were determined to move
beyond the traditional role of publishing opinions and extended to being
determined an underwriter their liability could become tied to any liabilities of
any other underwriter of the transaction.41
Current problems in the subprime market, the bankruptcies and therefore loss
of some of the former industry deep pockets, discussions in Congress of
assignee liability, and the myriad accounts of mortgage fraud on the part of
some lenders and some borrowers suggest that litigation against many of the
potentially participant parties will be among the possible responses to the
current crisis. With the majority of lower-rated RMBS tranches having been
bought, primarily by Collateralized Debt Obligations and subprime RMBS
exposures in CDOs averaging close to 45 percent4243 it is ironic that in a recent
hearing of the Senate Banking Committee titled Subprime Mortgage Market
Turmoil: Examining the Role of Securitization, there was no mention in the
written testimony by either Moodys or S&P, of this very profitable CDO
segment of their business.44
Perhaps another means to better define precisely the role CRAs play in some
structured-finance issues would be to consider their interactions with other
engaged parties who are also paid by the issuing client. In structured finance
transactions, legal isolation is usually a key consideration. Part of the rating
Source: Moodys Investors Service, Russian RMBS and Rating Process Overview (Oct. 4,
2006), available at http://www.ifc.org/ifcext/rpmmdp.nsf/AttachmentsByTitle/
Moody's+Presentation+1+Eng/$FILE/Microsoft+PowerPoint+-+1_Russian+RMBS+
Developments+(IFC+Moscow+Seminar).pdf.
47. BIS Study, supra note 13, at 49 (To avoid falling behind their peer group, they
then have incentives to trade with the crowd (reputation-based herding) to avoid the
reputational risk of acting differently from their peers (Scharfstein and Stein (1990)).
Similarly, herding can be compensation-based (Maug and Naik (1996)) when fee
contracts for delegated portfolio managers contain relative performance elements.
Second, institutions might infer private information about the quality of investments from
each others trades (as in Banerjee (1992) and Bikhchandani et al (1992)). Third,
institutional investors might trade together simply because they receive correlated private
information from analyzing the same indicators (Froot et al (1992)).).
48. JPMorgan, CDO Handbook 28 (May 29, 2001), available at
http://www.mathematik.uni-
ulm.de/finmath/ss_05/fe/JPMorganCDOHandbook.pdf. (See e.g S&P pioneered
market value CDO ratings in 1987 and cash flow ratings in 1988. By 1990, however,
Moodys dominated the rating of cash flow structures due to rating standards that more
flexibly addressed a wider range of portfolio credit quality and diversity. Beginning in
1996, Fitch began rating the second generation of market value CDOs. Moodys and S&P
came out with revised market value requirements in 1998 and 1999, respectively, and
became more active in that market.).
49. Frank Partnoy, How and Why Credit Rating Agencies Are Not Like Other
Gatekeepers 21, available at http://www.tcf.or.jp/data/20050928_Frank_Partnoy.pdf.
50. Authorite des Marches Financiers, supra note 31, at 10.
18 Joshua Rosner & Joseph R. Mason
time, compared to an average change of about 1.5 notches for corporate bonds.51
Having not been tested in a sustained period of economic volatility or economic
stress, it is unclear if the approaches employedwhich use computer simulations
to ascertain incidence of breaches in a tranches structure and not the severity of
those breaches52is an appropriate methodology.
Whether the rating of these reconfigured assets from sub-investment-grade
residential mortgage-backed securities to investment-grade tranches in CDOs is
the result of true diversification of risk or the result of rating agency alchemy is
unclear.53 What is clear is that the lack of liquidity, transparency, history and
available data coupled with unprecedented complexity has made it difficult for
all but the most well funded, well staffed and most sophisticated to analyze the
markets or assets. This has further increased market reliance on the CRAs.5455
This reliance is strikingly different from the traditional assets a rating agency
rates. Where a charter constrained investor disagrees with a rating agencys
assessment of a particular corporate issue it can avoid that issue and find other
issues within the desired asset class where it does agree with the rating. The more
active role of the rating agencies in structured finance, the essential oligopolistic
structure of the industry, the reality that their ratings are required for the sale of
investment grade structured finance securities and the theoretically consistent
application of their rating methods may mean that an investor who disagrees with
the rating agencies approach to structural risk rating is often precluded from any
investments in an entire asset class.56
51. John Ammer & Nathanael Clinton, Good News Is No News? The Impact of
Credit Rating Changes on the Pricing of Asset-Backed Securities (Board of Governors of
the Federal Reserve System International Finance Discussion Papers No. 809, Jul. 2004),
available at http://www.federalreserve.gov/pubs/ifdp/2004/809/ifdp809.htm#foot36.
52. Partnoy, supra note 49, at 22 (Although this process employs sophisticated
mathematical techniques, the conclusions can be somewhat dubious. For example, a
rating agency might run 100,000 computer simulations to determine the number of times
a breach would occur, that is, how often a particular tranche would lose value beyond a
certain level. However, the variable in this assessment is the number of breaches out of
the 100,000 runs, not the magnitude of the breach or any qualitative analysis of the
breach.).
53. Frank Partnoy & David A. Skeel Jr., The Promise and Perils of Credit
Derivatives 12 (University of Pennsylvania Law Sch. Working Paper No. 125, 2006),
available at http://lsr.nellco.org/cgi/viewcontent.cgi?article=1129&context=upenn/wps.
54. Authorite des Marches Financiers, supra note 31 (Due to the complexity of the
deals and the diversity of underlying assets involved, most investors (except investors in
junior or highly subordinated tranches) do not have dedicated credit analysts to properly
assess structured deals. As a result, many investors rely on rating agencies to assess the
quality of each structure, both at initiation and throughout the life of the transaction.).
55. BIS Study, supra note 13, at 44 (Overall, investors appeared to hold the view
that rating agencies are more important for structured finance than for traditional debt
instruments. Interviewees cited several factors at play: the rating agencies role in
modeling the risks of complex structured finance instruments; their key role in deal
structuring; and a clear information advantage, in particular over less sophisticated
investors. An important reason for the reliance on ratings seems to be that many client
mandates and internal investment guidelines prescribe minimum rating levels.).
56. SEC Hearings on Issues Relating to Credit Rating Agencies (Nov. 21, 2002)
(statement of Amy Lancellotta, Senior Counsel, Investment Company Institute, available
at http://ici.org/issues/dis/02_sec_2a-7_stmt.html [hereinafter Lancellotta Statement].
May 2007 Mortgage-Backed Security Ratings 19
57. Authorite des Marches Financiers, supra note 31, at 9. (See e.g Since 2005,
several rating agencies have proposed pricing services for the secondary market of
existing deals. To the extent that rating agencies are the sole decision makers for any
rating changes and due to the asymmetry of information between rating agencies and
market participants in this area where ratings constitute an essential if not exclusive basis
for setting a price, such a new service could raise questions about the information used by
rating agencies to provide market prices, and the mix of responsibilities this might create
between their role as rating entity and their presence in the secondary market for the
valuation of the rated paper.).
58. STRUCTURED CREDIT INVESTOR, Feb. 2, 2007, available at
http://www.structuredcreditinvestor.com/story.asp?PubID=250&ISS=22088&SID=15658
(People haven't addressed the market risk side of the deal.).
59. Id.
60. Role of Ratings, supra note 11, at 14.
20 Joshua Rosner & Joseph R. Mason
61. Rating Confusion, supra note 5 (For example, according to S&P, a rating of
"BBB" corresponds to a five- year default probability of 1.255% for asset backed
securities, but a higher default probability of 2.323% for corporate bonds. Likewise, a
rating of AA corresponds to a seven-year default probability of 0.315% for ABS,
0.420% for corporate bonds, and 0.701% for CDOs . At Moodys, municipal bond ratings
correspond to half the level of expected loss as corporate bond ratings for purposes of
rating CDOs. Inconsistent definitions make it hard for investors to use ratings to compare
the credit risk in different kinds of securities.).
62. Authorite des Marches Financiers, supra note 31, at 13. (See e.g In light of the
constant flow of new innovative structured products, questions should be raised at the
European level concerning the potential limitations of agencies' models - can everything
really be rated? and the meaning of ratings. This should be done in close cooperation
with the prudential authorities in charge of the implementation of Basle II and of the
OEEC regime, with the principal objective of encouraging understanding among the
participants and strengthening the market's credibility. It is also necessary to determine if
a structured finance rating is equivalent to a corporate rating and, within securitization
ratings, whether or not the rating of a cash CDO is equivalent to the rating of a synthetic
CDO. In this respect, is there a need for either differentiating structured finance and
corporate ratings or for requiring additional information regarding the volatility or
sensitivity attached to each structured finance rating, which, for certain products like
CDOs.)
63. Id.
64. COMMITTEE OF EUROPEAN BANKING SUPERVISORS, GUIDELINES ON THE
RECOGNITION OF EXTERNAL CREDIT ASSESSMENT INSTITUTIONS 1 (Jan. 20, 2006),
available at http://www.bundesbank.de/download/bankenaufsicht/pdf/cebs/GL07.pdf
[hereinafter Guidelines on External Credit].
65. See 12 C.F.R. 955.3(a), (b).
66. Letter from Stephen M. Cross, Director, Office of Supervision, Federal Housing
Finance Board, to Federal Home Loan Bank Presidents and Directors of Internal Audit
(Aug. 12, 2004), available at http://www.fhfb.gov/GetFile.aspx?FileID=68.
67. Guidelines on External Credit, supra note 64, at 161.
May 2007 Mortgage-Backed Security Ratings 21
68. Authorite des Marches Financiers, supra note 31. (See e.g. Due to pressure and
competition, rating agencies may be asked to rate complicated and innovative (though
well remunerated) deals within a very short time, which despite all the due process can
potentially lead to errors or an approximate assessment of the risks involved.)
69. Authorite des Marches Financiers, Ratings in the Securitization Industry 16 (Jan.
2006) [hereinafter Ratings in Securitization Industry].
70. Id.
71. Ratings in Securitization Industry, supra note 69 (The agencies claim that,
statistically, methodological changes rarely lead to rating changes. Investors say that
SPVs sometimes receive significant downgrades, but it is difficult to say whether these
are partly attributable to methodological changes or solely reflect a deterioration in the
underlying asset pool.).
72. Xudong Yongheng Deng & Anthony B. Sanders, Subordination Levels in
Structured Financing, Oct. 30, 2006, at 2, available at
http://fisher.osu.edu/~sanders_12/Sub.pdf.
73. Randall S. Krozner, Governor, North Carolina, Remarks to the 2007 Credit
Markets Symposium at the Charlotte Branch of the Federal Reserve: Recent Innovations
in Credit Markets (Mar. 22, 2007), available at
http://www.federalreserve.gov/BoardDocs/Speeches/2007/20070322/default.htm (Along
with liquidity, transparency in credit markets has also improved over time. Corporate
bond markets are more transparent thanks to a regulatory change that took effect in 2002.
Dealers must now report nearly all corporate bond trades to the NASD within fifteen
minutes, and the NASD immediately reports the trade data to the market. For asset-
backed securities and loans, price transparency is available from specialist vendors who
aggregate and disseminate dealers prices. For example, in the syndicated loan market,
one vendor currently aggregates data from more than seventy traders to price nearly
6,000 loans daily. Prices of many credit derivatives, including single-name CDS, credit
22 Joshua Rosner & Joseph R. Mason
derivative indexes, and credit index tranches, are widely available on services such as
Bloomberg or Reuters. Complex credit derivatives such as CDO tranches are an
exception to all this: They remain largely illiquid and nontransparent.).
74. BIS Study, supra note 13, at 39. (The quality and availability of surveillance
data during the life of a structure were generally felt to be inadequate, especially in
Europe, though improving rapidly. Investors often rely on quarterly reports by the trustee
and the rating agency. Sometimes these are sent to the initial investor only. A few
investors also pointed out that, as a result of their fee structures, rating agencies
sometimes cared less about surveillance than about initial ratings.).
75. Ratings in Securitization Industry, supra note 69, at 16 (The agencies rate a
transaction when it is carried out. Published changes in methodology apply to future
transactions. However, the agencies do not take the same approach when updating ratings
for past transactions. Some of the agencies do not revisit the old ratings but use the new
method in their surveillance. Thus, the surveillance process is based on both the old and
new methodologies.).
76. FitchRatings, Credit Policy Special Report 2007 Global Structured Finance
Outlook: Economic and Sector-by-Sector Analysis December 11 2006 at 19
77. Rating Stability of Fitch-Rated Global Cash Mezzanine Structured Finance
CDOs with Exposure to U.S. Subprime RMBS, Apr. 2, 2007, at 11.
May 2007 Mortgage-Backed Security Ratings 23
appropriate for risk-weighting analysis under Basel II.78 Market Value CDOs
are enjoying a revival with issuance more than doubling in 2006 from the year
before. MV CDOs appeal to managers because they offer greater trading
flexibility and can invest in a wide range of assets including high yield bonds and
bank loans and SF securities.79
In the wake of the recent credit problems in subprime residential mortgage
performance the rating agencies have issued research reports and held conference
calls with advise investors on the potential rating implications to both RMBS and
CDOs with subprime exposures. Many of the details of their analysis highlighted
potential risks to their assumptions.
In one of the more striking recent reports80 Moodys commented that it is
requesting increased levels of loan details from mortgage securitizers to enhance
their rating capabilities. The company stated the data fields essential for running
the model were established when the model was first introduced in 2002. Since
then, the mortgage market has evolved considerably, with the introduction of
many new products and an expansion of risks associated with them. The report
went on to point out that requests for data are broken into three categories:
The report implied that their current proprietary model for rating mortgage risk
does not currently incorporate the primary information they are now requesting
nor do they have the ability to assess the requested highly desirable or
desirable data to support their analysis. We again find reason to ask why
agencies are not expected to seek out more information than provided to them by
issuers or to verify even the non-financial data provided them by issuers.
Even in the existing data fields that the agency has used since 2002 as
primary inputs into their models they do not include important loan
information such as a borrowers debt-to-income (DTI), appraisal type and which
lender originated the loan.
Although the importance of DTI has been relatively less important in a
period of strong home price appreciation (HPA) it becomes of more importance
in a flat to declining HPA as borrowers have less access to appreciated equity
cushion in support of their ability to service a loan. The fact that this information
has not been a primary input to their model is even more surprising considering
that traditionally the loan to value (LTV) and FICO score and the borrowers DTI
are the three most significant measures of credit risk on a mortgage.81
Similarly, given the mortgage industrys move to automated appraisals the
lack of consideration of appraisal type as a primary input is concerning. Nonfull
appraisal techniques, such as AVMs, rely on public data that is ordinarily several
months old. In rising markets, AVMs depend on housing price data that is
slightly lower than current market conditions. However, in declining markets, the
AVM may overestimate property values given current market.82 The importance
of who the lender was, as it relates to loan quality, has also become obviously
more important in the wake of recent institutional failures of subprime lenders
many of whom failed due to weak underwriting standards.
The following figure details some of the more significant items of loan detail
that they are only now requesting.
* "When evaluating second lien loans, these fields are highly desirable for
the underlying first lien loans associated with these second lien loans"
81. Nomura Fixed Income Research, RMBS Basics 21 (Mar. 21, 2006), available at
http://www.securitization.net/pdf/Nomura/MBSBasics_31Mar06.pdf.
82. See, e.g., Jenine Fitter, Manufactured Housing: Waiting for the Rebound, 6
FITCHRATINGS, May 2004, at 1, available at
http://www.aaro.net/pdf/Reading%20Room/FitchReport.pdf (Fitch is concerned that,
under certain weakening housing conditions, any valuation method other than a full
appraisal is likely to overestimate property value. As it relates to automated valuation
models (AVMs), Fitch believes that the risk of property overvaluation is particularly
great in declining markets. In Fitchs opinion, such risk is paramount when alternative
valuation methods are employed because of the time lag in the underlying data collection
process. Nonfull appraisal techniques, such as AVMs, rely on public data that is
ordinarily several months old. In rising markets, AVMs depend on housing price data that
is slightly lower than current market conditions. However, in declining markets, the
AVM may overestimate property values given current market conditions.).
May 2007 Mortgage-Backed Security Ratings 25
Subprime Mortgage Distress Effect on CDOs,83 Fitch staff were asked about
the home price assumptions they are assuming. After several participants pressed
them on the issue a Fitch respondent stated that they assumed a mid-single digit
HPA. This is in stark contrast to fourth-quarter median home price data for 2005
and 2006 which confirms a national home price correction has been under way,
with the U.S. median home prices down 2.7%.84
Since mortgage backed securities primary risk exposures are to default and
prepayment risk, credit risk and market risk, many of the fields Moodys is
requesting have obvious and considerable importance in properly analyzing the
risks to RMBS and residential mortgage exposed CDOs. This would be
especially true in a period of weakening home prices, rising rates or declining
employment and income. Such evolutionary approaches to rating processes are
not rare.85
Part of the problem in the rating methods for ABS (including RMBS) and for
CDOs may well be that these methods, which are statistical, are built on other
rating models which are also statistical and the correlations between them may be
underappreciated in good times and rise suddenly in bad times.
Default correlations are higher during economic downturns or recessions and
it has been viewed that macro-economic factors are the main driver of portfolio
losses86 and at the portfolio level, dependencies between defaults are crucial and
little is known about them.87 As agency models are increasingly used to rate
assets such as RMBS and CDOs that themselves are more directly tied to
exogenous macro-economic conditions than by the endogenous conditions of a
particular enterprise, such as a corporate issuer, the importance of measurement
data to inform correlation modeling should be more significant in these assets.
Historic data is used by agencies in informing correlation assumptions and they
advocate the use of empirical default correlation to benchmark internal
models.88
83. Fitch Webcast: Subprime Mortgage Distress Effect on CDOs Tomorrow 09:30
am HK/SG Standard Time (Apr. 22, 2007), available at
http://www.derivativefitch.com/press_release_frame.cfm?pr_id=353234.
84. Fitch Reports on How the Housing and Mortgage Market Downturn Could
Affect Municipal Credit, BUSINESS WIRE, Apr. 25, 2007, available at
http://home.businesswire.com/portal/site/google/index.jsp?ndmViewId=news_view&new
sId=20070425006074&newsLang=en.
85. Jody Shenn, BLOOMBERG, Apr. 4, 2007, available at
http://www.elitetrader.com/vb/showthread.php?threadid=91393. (On April 4, 2007
Bloomberg reported Moodys (is) concerned that the growth of synthetics, or credit
swaps, may leave more CDOs invested in other CDOs exposed to the same bonds as they
are. The company said its models were developed using the data that was available at the
time, such as transactions backed by cash collateral. Moodys is now working on a
research project to reassess the correlation between CDOs at time when exposures can be
infinitely replicated, it said. Moodys sees increasing correlations in performance,
which suggests it will require more protection for bondholders when the project is
finished. Well update the market with our final findings when were done, Fu said.).
86. Arnaud de Servigny & Olivier Renault, Default Correlation: Empirical
Evidence, STANDARD & POORS, Nov. 23, 2002, at 12, available at http://www.risklab-
madrid.uam.es/es/jornadas/2002/CorrelationPresentation.pdf.
87. Id. at 3.
88. Id. at 28.
26 Joshua Rosner & Joseph R. Mason
Given that there is little empirical data on many of these newer assets to be
rated, such as non-traditional mortgage assets, should we to expect that their
correlation assumptions within these newer assets will be subject to increased
volatility as down-cycle empirical data is being captured or should we assume
that the statistical assumptions they use will themselves be conservative enough
in their adjustments to compensate for the scarcity of full-cycle data?
As example, one of the key considerations in RMBS ratings is the servicer
rating of the firm(s) servicing the pool. When home prices are rising,
employment is strong, and interest rates are benign, mortgage performance tends
to be strong. Such an environment leads to low servicing costs, the management
of servicers being less strained, capital requirements and liquidity risks of
servicers being low, and the sophistication and operational efficiencies of
servicers being less important. As a result servicers ratings are less subject to the
rigors of market stress testing since almost everyone appears to be doing a decent
job of servicing.
As problems emerge for mortgage performance, such as reduced home price
appreciation or rising interest rates or declining incomes or employment, market
liquidity becomes more scarce as investors become increasingly risk averse. Just
as the relation of this effect is positively correlated to servicing performance on
the upside it is positively correlated to servicing on the downside. This places the
value of using servicer ratings as a forecasting tool in RMBS performance at risk
since, at the time mortgage performance declines the cost to servicers of
servicing problem loans rise as do the demands on their operational platforms
efficiency and management. So, as liquidity may be moving away the mortgage
space it will also be moving away from the servicer market at precisely the time
it is most dear.
In an environment of stabilizing home prices or a slowly increasing rate
environment, assuming other aspects of the mortgage process were effective
(underwriting quality as example), the agencies models might not be subject to
stress. In an environment where market liquidity reverses quickly both RMBS
and servicer rating models would be subject to significant and perhaps
unanticipated stress that the agencies only catch after the effect is generally seen.
This raises further question about the predictive value of their models in
structured finance.
In a real world example, on October 30, 2006 Fitch announced that it upped
the residential primary servicer rating of New Century from RPS3+ from RPS3
(the ratings are based on a scale of 1-5 with 1 being highest). At that time the
Companys servicing operation was commended for its competent management
team, established servicing platform, capable default technology, and enhanced
cross-functional training platforms.89
At that time, although the environment in which New Century was more
challenging that in recent prior years, New Century appeared to be managing
well. The Company, which operated three businesses; portfolio, mortgage loan
operations and servicing stated their average workforce increased from 5,594
for the nine months ended September 30, 2005 to 7,119 for the nine months
ended September 30, 2006, an increase of 27.3 percent. This increase in
89. See New Century Servicing Improves Fitch Bumps Subprime Servicer
Rating, MORTGAGE DAILY, Oct. 30, 2006, available at
http://www.mortgagedaily.com/NewCenturyRating103006.asp.
May 2007 Mortgage-Backed Security Ratings 27
workforce was mainly due to our acquisition of the mortgage loan origination
platform of RBC Mortgage in September 2005. The remainder of the increase
was primarily due to growth in our servicing platform and the mortgage loan
portfolio90. In the same filing the Company stated its Servicing income
increased 100.8 percent to $47.4 million for the nine months ended
September 30, 2006, compared to $23.6 million for the same period in 2005.
Unfortunately, instead of filing a timely Annual Report for 2006, on March 3,
2007 the Company announced a late filing91 and warned it will restate its
consolidated financial results for the quarters ended March 31, June 30 and
September 30, 2006 and that it expected to ultimately report a pre-tax loss for
both the fourth quarter and full-year of 2006.
While the rating agencies could not be expected to have known of the
corporate accounting issues of New Century and the Company, in its statement,
indicated that other business lines than the servicing business were primarily
responsible for the losses, the liquidity problems that ensued ultimately could
have impaired the Companys servicing of loans. As a result, on March 5,
Moodys downgraded New Centurys servicer rating from SQ3+ to SQ4.92 On
March 7, Fitch lowered the ratings to RPS4 from RPS3+.93 Shortly after, New
Century defaulted on repurchase obligations. This resulted, in part, from
downgrades of their servicer ratings.94 On April 6, Moodys downgraded the
servicer rating of New Century to SQ4- from SQ4.95
So, in building models, the agencies might build a servicer rating based on
the performance of a servicer whose business may become at risk based on
businesses other than servicing. Moreover, as the mortgage market becomes
more volatile and difficult for lenders whose loans may be those serviced by the
same company, the risks are compounded. To then use the servicer rating as part
of the rating process for a mortgage-backed security can have the mathematically
equivalent impact of double counting. Obviously, the rating agencies role itself
can define the difference between life and death.
Just as one of the key considerations in the evaluation of RMBS assets is the
servicing entity, an important pillar of Fitchs rating process for a CDO is an
90. New Century Financial Corp., Quarterly Report (Form 10-Q), at 62 (Sept. 30,
2006), available at http://www.sec.gov/Archives/edgar/data/1287286/
000089256906001359/a24944e10vq.htm#103.
91. New Century Financial Corp., Notification of Late Filing (Form 12b-25),
available at http://www.sec.gov/Archives/edgar/data/1287286/
000119312507045551/dnt10k.htm.
92. New Century Ratings Downgraded Moody's lowers servicer rating to SQ4,
MORTGAGE DAILY, available at
http://www.mortgagedaily.com/NewCenturyRating030507.asp.
93. See New Century's Survival in Doubt Fitch lowers servicer rating,
MORTGAGE DAILY, Mar. 7, 2007, available at
http://www.mortgagedaily.com/NewCenturyRating030707.asp.
94. New Century Crumbling Defaults disclosed in SEC filing, MORTGAGE DAILY,
available at http://www.mortgagedaily.com/NewCenturyCreditors031207.asp.
95. Coco Salazar, Mergers Dissolve: Latest Mortgage M&As Corp activity,
MORTGAGE DAILY, Apr. 6, 2007, available at
http://www.mortgagedaily.com/Mergers040607.asp.
28 Joshua Rosner & Joseph R. Mason
E. Regulatory Issues
Given the rapid growth in issuance shown in the table above it would seem
appropriate to consider whether the rating agencies have been appropriately
spending on operational risk issues including systems and staff, relative to the
growth in rated assets.100
Just as bank regulators are tasked with examining regulated institutions and
look at such issues as capital, assets, management, earnings, liquidity and
sensitivity to market risk, the SEC staff review aspects of the rating agencies
business and seem to have some ability to, among other powers, make sure that
the agencies are staffed with an appropriate number of experienced ratings staff
and operate independently of the companies they rate.101 The SEC examines the
rating agencies every five years102 and their authorities, in approval of new and
continuing registration of NRSROs, are defined by law.
Last year, Congress passed the Credit Rating Agency Reform Act of
2006103 which, among other things authorized the SEC to suspend or revoke the
NRSRO status of a current registrant if it didnt have the financial or managerial
resources to produce ratings with integrity.104 In a letter to SEC Secretary Katz in
June of 2005, S&Ps President Kathleen Corbet in answer to the SECs question:
The Commission requests comment on the appropriate subjective criteria that a
credit rating agency should use in assessing the experience and training of an
analyst to meet the proposed NRSRO definition? responded that While Ratings
Services is confident that its standards and procedures for analyst background
and training would meet any minimum requirements imposed by the
Commission, Ratings Services does not support the promulgation of NRSRO
designation criteria that are conditioned on specific attributes of a rating agencys
staff.105
of credit rating are human resources and the skills of the staff employed. The professional
background and qualification of the CRAs employees is not published and there is not a
framework of statutory rules requiring a specific professional standard of training. So it
concludes that it is difficult to assess whether the resources employed by the CRAs are
sufficient.
Similarly, Rating Evidence stresses that only a very few analysts undergo training to
absorb the CRAs rating philosophies and approaches at one of the universities offering
rating education. Thus usually the analysts receive only on-the-job and in-house training.
This respondent concurs that the confidence in the ratings would increase if rating
analysts education would be more structured like in other professions.).
101. See, e.g. Credit Rating Agency Reform Act of 2006, S. 3850, 109th Cong.
2(E) (2006), available at http://frwebgate.access.gpo.gov/cgi-
bin/getdoc.cgi?dbname=109_cong_bills&docid=f:s3850enr.txt.pdf.
102. See Lancellotta Statement, supra note 56.
103. Credit Rating Agency Reform Act, supra note 101.
104. Id. (The Commission shall grant registration (to applicants for NRSRO
status) under this subsection (i) if the Commission finds that the requirements of this
section are satisfied; and (ii) unless the Commission finds (in which case the Commission
shall deny such registration) that (I) the applicant does not have adequate financial and
managerial resources to consistently produce credit ratings with integrity and to
materially comply with the procedures and methodologies disclosed under paragraph
(1)(B) and with sub-sections (g), (h), (i), and (j); or (II) if the applicant were so
registered, its registration would be subject to suspension or revocation under subsection
(d).).
105. Letter from Kathleen A. Corbet, President, Standard & Poors, to Jonathan G.
Katz, Secretary, United States Securities and Exchange Commission 8 (June 9, 2005),
available at http://www.sec.gov/rules/proposed/s70405/standardpoors060905.pdf.
30 Joshua Rosner & Joseph R. Mason
106. Id. at 7. (Ratings Services believes that the Commission could appropriately
consider the number of analysts employed by a credit rating agency in the NRSRO
designation process. However, mandating disclosure by rating agencies of the average
number of issues rated by those analysts would serve little practical purpose and might in
fact mislead the market.).
107. Structured Finance Roles, FIN. ENGINEERING NEWS (2005), available at
http://www.fenews.com/fen43/euro_angles/euro_angles.html.
108. Before the S. Comm on Banking, Housing, and Urban Affairs, 109th Cong. 1-
2 (2005) (statement of Stephen W. Joynt, President and Chief Executive Officer of Fitch
Ratings), available at http://banking.senate.gov/_files/joynt.pdf.
109. Before the S. Comm on Banking, Housing, and Urban Affairs, 109th Cong. 1-
2 (2005) (statement of Raymond W. McDaniel, President, Moodys Investors Service),
available at http://banking.senate.gov/_files/mcdaniel.pdf.
110. Press Release, International Organization of Securities Commissions, Iosco
Releases Code Of Conduct Fundamentals For Credit Rating Agencies (Dec. 23, 2004)
available at http://www.fsa.go.jp/inter/ios/f-20041224-3/02.pdf.
111. See Securities and Exchange Commission Release34-55231 File No. S7-04-07
Oversight of Credit Rating Agencies Registered as Nationally Recognized Statistical
Rating Organizations p.10 SECURITIES AND EXCHANGE COMMISSION 17
CFR Parts 240 and 249b [Release No. 34-55231; File No. S7-04-07] RIN 3235-AJ78.
May 2007 Mortgage-Backed Security Ratings 31
Over the past several years we have witnessed a seemingly significant public
role of the agencies as described by the de-facto designation they have received
as gatekeepers of bank capital adequacy.
There are examples over the past few years that, given their increased
importance to global capital flows, are worth public policy consideration.
Whether or not these activities are a result of real or perceived conflicts of
interests on behalf of issuing clients is unclear. What is clear is that NRSRO
powers have extended to areas of public policy in ways we have not witnessed
before.
In early 2004, after accounting problems were discovered at Freddie Mac but
before those of Fannie Mae were fully uncovered, Congress again embarked on a
legislative process to create a new regulator with enhanced powers. One of the
key provisions legislators considered was one that would better define the
receivership authority of the GSEs regulator in case they became seriously
undercapitalized. In early April S&P hinted about a possible downgrade of GSE
debt if a new regulator had receivership powers.114 This announcement
supported the GSEs goals of trying to prevent receivership authority from being
included in legislation. It is unclear if this was done out of support for one of the
115. Bara Vaida, Taking Cover, NATL J., May 28, 2005, available at
http://w4.stern.nyu.edu/news/news.cfm?doc_id=4544.
116. Robert A. Eisenbeis, W. Scott Frame & Larry D. Wall, Resolving Large
Financial Intermediaries: Banks Versus Housing Enterprises (Federal Reserve Bank of
Atlanta, Working Paper Series, Paper No. 2004-23a, 2004), available at
http://www.frbatlanta.org/filelegacydocs/wp0423a.pdf.
117. Examining the Role of Credit Rating Agencies in Capital Markets: Hearing
Before the S. Comm. on Banking, Housing, and Urban Affairs, 109th Cong. 21 (2005),
available at http://a257.g.akamaitech.net/7/257/2422/19jul20061200/
www.access.gpo.gov/congress/senate/pdf/109hrg/28059.pdf.
May 2007 Mortgage-Backed Security Ratings 33
motion though it did not rule whether or not the securitizations were true sales.
Although this case could have caused the rating agencies to take the same
position as the Georgia law, of ambiguity making it difficult to rate the risks to
noteholders they chose not to. In fact, one of the agencies appeared to pressure
attorneys to avoid commenting on the matter in legal opinions. Standard &
Poor's insisted that attorneys submitting true-sale opinions to the rating agency
stop referring to LTV, noting that the court never made a final decision and that
such citations inappropriately cast doubt on the opinion. Seven months later, in a
delicately worded press release, S&P withdrew that prohibition--apparently
because lawyers refused to ignore such an obvious legal land mine.121
Rating agency claims of their inability to assess the limits of credit risk
exposures to investors in Georgia as a result of the 2002 legislation becomes
more interesting to note given their relative silence in addressing potential
ongoing risks to investors that could be posed by the explicit authority of the
FDIC, after meeting certain thresholds, to abrogate contracts as receiver of a
failing institution.122 In the case of NextBank the FDIC abrogated an accelerated
payment contract provision for noteholders of a credit card securitization. Unlike
the LTV Steel case, there was a clear determination in Bank of New York v.
FDIC. Although the FDIC did not seek to invalidate or question the legal
isolation of the securitization trust it did, nonetheless, abrogate certain payments
to noteholders. It is therefore not only unclear how far the FDICs authority
would extend into a securitization trust but it is unclear if this is considered in
rating assessments.
Corporate bond values are not dependent upon recourse, servicing incentives,
and other external factors. That is because even if the bonds are collateralized by
specific assets, the corporation is explicitly tied to those assets and services and
otherwise maintains their profitability within the corporate structure. If the
collateral backing the corporate bonds experience difficulty generating cash
flows sufficient to pay creditors, the corporation can channel excess revenues
from other operations to cover the shortfall. Those differences, described below,
create the basis for important fundamental differences in rating corporate bonds
and RMBS or ABS.
121. Tim Reason, False security? Corporate insolvencies are testing whether
securitization is a stable structure or a flimsy faade- Bankruptcy 2003, CFO: MAGAZINE
FOR SENIOR FINANCIAL EXECUTIVES, June 2003.
122. The Bank of New York v. Federal Deposit Insurance Company, 453
F.Supp.2d 82 (D.D.C. 2006), available at https://ecf.dcd.uscourts.gov/cgi-
bin/show_public_doc?2003cv1221-75.
May 2007 Mortgage-Backed Security Ratings 35
prima facie evidence to argue that traditional ratings practices create dire
consequences when applied to RMBS and CDOs (and other structured finance
securities) in fundamental structural differences between the different firms
involved, i.e., the RMBS pool trust versus the corporation. We demonstrate
below that merely creating a pool of investments does not build diversification.
Then we show that the static nature of the firm in structured finance makes the
securities of that firm much more volatile than equivalently-rated corporate
securities. When those static firms invest solely in fixed-income investments
that is no reason to delay downgrades as there is no way for the static firm
investments to recover and no way to redeploy capital. Since there is also no way
for the firm to siphon off excess earnings, however, there is also no reason to
delay upgrades.
Hence, RMBS, as part of a static capital structure, funding a static investment
strategy composed exclusively of fixed income assets, and trading thinly over-
the-counter with few market price signals, require dynamic ratings in order to
adequately reflect risk. The main point is therefore that, in the case of RMBS,
bond ratings are being used for something they were not initially intended and
often sold on the basis of a fundamental misunderstanding of diversification.
Below we show the implications of these fundamental shortcomings. Section
III.A focuses on how the principal of diversification has been misrepresented by
Wall Street. Section III.B analyzes the implications of rating RMBS with
corporate debt rating techniques. Section III.C provides several case examples to
illustrate common performance characteristics of ABS and RMBS.
One of the primary myths perpetuated on Wall Street is that mortgage pools
are diversified. Commonly cited claims read like the Wall Street Journal editorial
on April 25, 2007, in which Ted Frank wrote that, diversification [in
mortgage pools] reduces the risk of lending to borrowers with suboptimal credit.
Such claims represent a fundamental misunderstanding of the distinction between
risk pooling, risk sharing, and diversification.
Risk pooling is the mere act of adding another investment to build up a
portfolio. Suppose we have one $100,000 mortgage with a 2 percent probability
of complete loss (the borrower never pays and an earthquake sucks up the house
and land). Expected payout on the loan is $98,000 and standard deviation (risk)
of the investment is 14 percent. It seems like we can lower the standard deviation
of the outcome by accumulating a pool of mortgages. By accumulating, say,
10,000 identical uncorrelated mortgages the standard deviation of the pool indeed
declines to 14%/10,000 = 0.14%.
The problem is that we have decreased the percent of portfolio risk (standard
deviation) but increased the dollar amount of the portfolio. Basic MBA-level
finance, however, teaches that if we wish to compare the risk of two investments
of different scale, we must compare the dollar profits from the two. When we
combine a number of uncorrelated investments both expected profit and standard
deviation grow in direct proportion to the number of investments. Since the
36 Joshua Rosner & Joseph R. Mason
risk/return tradeoff does not improve with the accumulation of more mortgages,
there is no diversification in pooling mortgages.123
The advantage of pooling investments lies in the ability to more efficiently
share risk with other investors. After the pool is constructed, the risk of the
mortgage pool can be distributed by allowing investors to share the risk through
purchasing securities representing a partial claim on the pool. Those investors,
each with a fixed-size portfolio to allocate across investments, can now allocate
risk across more sources of risk, helping them diversify their own portfolios.
According to Bodie, Kane, and Marcus, portfolio risk management is about the
allocation of a fixed investment budget to assets that are not perfectly correlated.
In this environment, rate of return statistics, that is expected returns, variances,
and covariances are sufficient to optimize the investment portfolio. Choices
among alternative investments of a different magnitude require that we abandon
rates of return in favor of dollar profits.124
123. See ZVI BODIE, ALEX KANE & ALAN J. MARCUS, INVESTMENTS 232-233 (7th
ed., McGraw Hill/Irwin 2007).
124. Id. at 234.
May 2007 Mortgage-Backed Security Ratings 37
In contrast, the assets in the mortgage pool are a pre-specified pool of fixed-
income investments. While fixed-income investments may outperform
expectations, that only means they pay in full as originally contracted. The upside
potential is therefore the expected loss rate, which has historically been just over
0.15 percent in the mortgage sector.125 Since the mortgage pool must be brain-
dead in order to avoid being classified as an investment company (and therefore
taxed) the assets are pre-specified. The pool trustee therefore has no authority to
change the pool investment strategy in reaction to unexpected underperformance.
Since the pool trust is brain-dead, the pool trustee also has no authority to change
the capital structure by selling new equity to buffer creditors from the increased
risk. The pools ability to accumulate additional equity by saving current period
excess earnings is inherently limited by the minimal upside potential on the
fixed-income assets.
These simple distinctions create several important differences between
corporate debt and mortgage pool debt (RMBS) with respect to 1) the
fundamental source of default risk; 2) the meaning and relevance of performance
point estimates; 3) the dynamics of cumulative losses; 4) the shape of statistical
distributions on conditional losses; 5) the width (standard error) of those
statistical distributions; and ultimately 6) the nature of decision-making in
upgrade and downgrade decisions and 7) the use and relevance of credit
enhancement.
125. The 1-4 family first-lien home mortgage chargeoff rate at all commercial
banks averaged 0.168% across the 1992-2006 period and never exceeded 0.348%
annually. See Statistics on Banking, available at www.fdic.gov.
38 Joshua Rosner & Joseph R. Mason
cannot redirect the investments (although the servicer can alter the servicing
platform ex post, which is precisely the point of the structural changes outlined in
our earlier paper).
There is a substantial risk however, that the static mortgage pool investments
will be imperceptibly fundamentally flawed ex ante, since changes to origination
standards may be manifested in adverse loan performance only after a substantial
lag. Of course, such a movement precludes future securitizations from that
originator due to the reputational cost, but the originator also will not have to
provide costly ex post support to preserve market access. As the saying goes, the
only securitization without recourse is the last. Furthermore, static mortgage
pools offer no opportunity for vintage diversification, save through master trust
structures, which are cost effective only for the largest issuers. Even in a master
trust, however, vintage diversification can only occur within the mortgage sector:
the mortgage pool cannot accumulate non-mortgage investments.
In summary, an investment in corporate debt is an investment in a proven
ongoing dynamic investment strategy. An investment in a mortgage pool is an
investment in a static pool of unproven fixed-income investments. We explain
below how those differences make rating corporate debt very different from
rating RMBS.
2. Performance Point Estimates that are the Basis for Corporate Bond
Ratings are not Relevant for RMBS
Corporate Investments
Mortgage Pool
Expected loss point estimates are more difficult to compute for mortgage
pools because mortgage pools are initially relatively unseasoned.126 Figure 11
illustrates the difference between loss estimates on corporate investments and
mortgage pools. The steep rise in losses in the left-most region of the expected
loss curve for the mortgage pool characterizes the ramp-up problem for
unseasoned investments.
A steep path of cumulative losses during the ramp-up period can indicate a
problem with the underlying collateral. The problem becomes, how do you
identify too steep? Uncertainty about the steepness of the ramp-up period has
therefore been the source of most previous private structured finance problems.
When pools have shown higher defaults during the ramp-up phase, the
mortgage pools have not be able to trap adequate excess spread to fund the credit
enhancement (which provides the primary backstop for default). This was a
typical problem with deals in the early 1990s, when several deals had to be
restructured in order to avoid complete failure.127
More recently, many mortgage pools exhibited similar early payment default
(EPD) problems where defaults rose faster than anticipated while the pool was
seasoning. On April 17, 2007, Wells Fargo announced that its revenue was
reduced by $90 million because of steps it took to address continued woes in the
subprime mortgage sector. The bank reduced the value of subprime loans in its
portfolio and set aside more funds for early payment default on loans that have
been securitized.128 After studying the collateral attributes of early payment
default (EPD) loans and comparing them to loans that did not default in the first
12 months after issuance, Fitch found that Fair Isaac Corp. (FICO) scores have
become less significant as an early default indicator when other high risk loan
attributes, such as piggyback second liens or loans with no-income verification,
are present.129 Such conclusions support our earlier assertions that fundamental
changes to origination and servicing standards over the past decade have led to
difficulties in valuing and pricing RMBS.
Given that problems predicting a moving expected loss rate early in the life
of a mortgage pool are the source of most RMBS problems, it appears that
predicting a moving expected loss rate on unseasoned collateral is much more
difficult than predicting a constant expected loss rate on an ongoing established
corporate investment strategy.
The point was made earlier that corporate obligations invest in real assets. As
a result, corporate investment portfolios can substantially outperform or
underperform expectations. Furthermore, corporate investment strategies can be
altered to take advantage of new investment opportunities, sometimes generating
returns that offset earlier losses.
Mortgage pools invest in fixed-income assets. Fixed-income assets cannot
substantially outperform expectations, but can significantly underperform
expectations. Furthermore, since the investment strategy and capital structure of
the mortgage pool is static, significant losses that accumulate in the pool cannot
be offset by either significant outperformance (that would generate excess
retained earnings) or changes to the investment strategy. The dynamics arising
from the different nature of the underlying collateral in corporate and mortgage
pool investments is illustrated in Figure 12. The solid s-shaped cumulative loss
profile (in gray) is that for a dynamic pool of real corporate investments, which
can recover from adverse shocks due to the real nature of the assets.
In contrast, because some corporate investments may be written off as losses
while others may perform substantially beyond expectations, corporate debt may
illustrate fluctuating cumulative loss dynamics.
Because mortgage pools are invested exclusively in fixed-income with little
potential for substantial outperformance, however, the pool has little or no
opportunity to recover from adverse loss rates. Static mortgage pools, because of
their ramp-up period and their limited upside potential, therefore illustrate rising
cumulative loss dynamics, with deviations lying primarily in the steepness of the
ramp-up curve.
129. Fitch Report Details Drivers of Early U.S. Subprime Defaults, BUSINESS
WIRE, Apr. 16, 2007.
May 2007 Mortgage-Backed Security Ratings 41
Corporate Investments
% Loss on Asset Pool
Expected
Actual
Mortgage Pool
The rising cumulative loss profile (in black) is that for the mortgage pool,
which is a static pool of fixed-income investments. Note that the solid black line,
lying only slightly below the expected loss profile (the dashed black line),
represents the best possible outcome given previous losses, i.e., that in which the
remaining fixed-income investments in the pool pay out all remaining principal
and interest. The small amount of equity in the pool relative to the amount of
debt securities dictates that the best case line lie only a small distance below
the cumulative expected loss profile.130
While corporate investments have equal upside and downside performance
risk, mortgage pool investments have more downside risk than upside. That
means that the shape of the statistical distributions used to calculate expected
default probabilities, that is, bond ratings, is significantly skewed.
130. While some may be tempted to argue that the master trust structure (allowing
issuers other pools to co-insure one another) is meant to address that problem, it is still
the case, however, that pool cumulative loss profile will not recover. Other pools merely
contribute their own equity to subsidize non-performing pool.
42 Joshua Rosner & Joseph R. Mason
The rightmost panel of Figure 13 illustrates further, however, that the skewed
mortgage pool loss distribution is centered around a moving point estimate,
whereas the (normal) corporate investment loss distribution is centered around a
constant point estimate. The ratings exercise for RMBS therefore becomes
estimating the area under an unknown distributional shape with a moving center.
But there is more.
The mortgage pools ability to accumulate additional equity by saving
current period excess earnings is inherently limited by the minimal upside
potential on the fixed-income assets. Furthermore, since the pool is brain-dead,
the pool trustee has no authority to change their capital structure by selling new
equity to buffer creditors from that increased risk. A mortgage pools investment
cash flows are therefore constant at best, and typically deteriorating to some
extent as losses accumulate (at a decreasing rate) through the life of the pool.
Residual equity is built up only from cash flows that accrue above
expectations, that is, unexpectedly low default or prepayment rates. With low
expected loss rates to begin with, upward performance potential is typically
limited to the two percent or so of equity typically included in RMBS structures.
Both corporate investments and mortgage pools, however, face equal
downside potential of total loss. Since there is limited upside potential to the
point estimate losses for the mortgage pool, the distribution of probable point
estimates in any time period is skewed rather than normally distributed. Figure
14 confirms the general shape of the skewed distribution from FitchIBCAs
Monte Carlo simulation of residential mortgage pool losses presented in their
technical literature.
May 2007 Mortgage-Backed Security Ratings 43
Source: FitchIBCA, ResiLogic: U.S. Residential Mortgage Loss Model Technical Document, Jan.
18, 2007 at 11.
% Loss on Asset Pool
The dashed black line in Figure 15 is the expected loss profile for the
mortgage pool. The skewed distribution along the dashed line represents the
statistical distribution of expected loss outcomes for the mortgage pool,
conditional upon the specific time period. The solid black line is actual mortgage
pool performance. The solid black lines position above the dashed line
represents substantial performance deterioration, in that actual losses are larger
than expected losses. The downgrade decision therefore centers on the question,
Are actual losses severe enough to represent an abnormal deviation from
performance substantial enough to justify a securities downgrade? Substantial
enough in this context is usually taken to mean something akin to two or more
standard deviations away from the mean.
May 2007 Mortgage-Backed Security Ratings 45
The gray dashed line is the midpoint of the expected loss distribution for
corporate investments. The gray normal distribution is the distribution on
expected losses on corporate investment. Note that in the middle of the life of the
mortgage pool, the actual losses on the mortgage pool are not two standard
deviations away from the mean of the corporate loss distribution (point 1 in
Figure 15). That loss level, however, is already well into the tail of the skewed
(proper) mortgage pool loss distribution (point 2 in Figure 15), however,
suggesting that the RMBS should have been downgraded some time ago.
Standard normally-distributed loss assumptions used for corporate debt will
therefore downgrade RMBS much later than is warranted under accurate
distributional assumptions.
Additionally, since the skewed loss distributions for mortgage pool
performance are narrowing over time and the static fixed-income portfolio has no
chance of significant recovery, the RMBS have no chance of later upgrade. When
mortgage pools do not perform, therefore, the outcome is dramatic and final.
% Loss on Asset Pool
Credit Enhancement
Cumulative Losses Loss Boundary
% Loss on Asset Pool
As time proceeds the cumulative loss distribution tightens around the mean.
Hence, in the event of positive performance the deal gets safer and saferthat is,
the amount of the shaded area under the statistical distribution that lies above the
loss boundary indicating credit enhancement coverage gets smaller and smaller.
Moving right across Figure 17, the level of loss that credit enhancement protects
against becomes increasingly unlikely. The expense of maintaining coverage
against such a level of loss exposure is therefore wasted.
8. Summary
Case studies can illustrate the phenomena described in the previous section.
We use ABS*TRAK to highlight how to identify key movements in pools and
securities performance relevant to the conclusions above. ABS*TRAK uses an
automated deal-by-deal data upload with precise transaction structures manually
coded. ABS*TRAK has distinct calibration and monitoring phases. In the
calibration phase, outputs on each transaction are scaled so the ABS*TRAK on
the most senior class of notes matches precisely the official rating agency rating.
Calibration insures internal consistency across time for all classes of notes
analyzed by the ABS*TRAK. In the monitoring phase, updates are generated
from monthly servicing data. Numeric outputs include the ABS*TRAK and the
theoretical credit curve, also called theoretical ABS*TRAK, which are
presented in the figures that follow as T.ABS*TRAK1. Since it is the
representative dynamics of the deal that are of interest, rather than the specific
deals themselves, the examples that follow maintain the issuers anonymity.
Figure 18 illustrates pool dynamics that should never legally occur. Merely
knowing that defaults cannot recover in a static pool, the downward spike in pool
balance is a red flag that something is amiss. The downward spike in annualized
periodic defaults suggests that perhaps some new accounts were added to
sweeten the credit quality. None of those dynamics, however, show up in the
simple cumulative default curve, however, because the trustee is managing the
pool to wash out the types of dynamics that are typically monitored by investors.
As an investor, however, the potential for pool management creates a legal
risk in addition to the fundamental (and obviously misunderstood) collateral risk.
Investors make the right choice here. Investors in the AA tranche bid up the
spread on the securities even though the loss dynamics were theoretically saved
by the trustees action (i.e., the red line remains a zero spread). Note that the y-
axis measures basis point discount on the security, so that 10,000 bps is a total
loss. The AA-rated A-1 tranche investors therefore face a near-total loss on
investment in this deal
Investors in the BBB-rated B-1 tranche act faster, and investors exposed to
the BB-rated C-1 tranche even faster still. Note that none of the prices recover
afterward, as expected by the analysis of the previous section. Both the B-1 and
C-1 investors face a total loss scenario, although the C-1 is a total loss at about
month 5 whereas the B-1 is a total loss at about month 24.
May 2007 Mortgage-Backed Security Ratings 49
The example in Figure 19 shows what happens when defaults start becoming
fundamentally more complex. The annualized periodic default curve in this case
looks smooth, so there is not the previous worry about some kind of pool bailout
or trustee risk. Nonetheless, losses are mounting faster than anticipated. Evidence
for that assertion shows up in the cumulative default profile, which is still not
turning downward after 10 months.
What happened here was that bankruptcies hit the pool after funding so that
pool losses occurred sooner than expected. In essence, the deal was mis-
engineered in that the loss dynamics were miscalculated. This is a good example
of the ramp-up problem presented above.
Eventually, the actual cumulative loss profile catches up with historical
expectations, primarily due to the slowdown in annualized periodic defaults in
May and June. That realignment is reflected in a recovery in AAA-rated tranche
prices at about month 11. Prior to that date, investors were deeply discounting the
price due to concerns about the fundamental pool performance. Similarly,
investors in the A-rated tranche also regain confidence in month 12 and the
pricing of the B-1 tranche returns to normal at that point.
50 Joshua Rosner & Joseph R. Mason
investors would really like to know what is going on here. There seems to be
something abnormal in such quick movements.
In this case the cumulative default curve may be beginning to level off after
30 months, but we cannot be sure that is due to fundamentals because of the
spike and recovery in the annualized periodic defaults. Investors may not be sure
what is going on either. Investors in the AAA-rated B-1 tranche initially
discounted their exposures but then priced the securities well below relevant
historical risk exposures based on the ABS*TRAK model (the red line), which
shows a lot of residual risk in their exposure. Investors in the AAA-rated B-3
tranche are far less sanguine. Interestingly, even though those investors are lower
in the waterfall, they initially discounted the securities by less than the B-1
investors. The B-3 investors, however, are discounting the securities much later
in the life of the deal, when the loss dynamics are getting strange. The B-3
investors are therefore pricing risk, while the B-1 investors are not.
These examples show the sensitivity of RMBS prices to how well loss
assumptions hold up during the very uncertain early seasoning stages of the deal.
If the assumptions hold, the deal will perform well up to maturity. If not, the deal
will fall apart.
In summary, if capital structure and investment decisions can be changed to
ameliorate new risk (and the firm is willing to do so), it is fine for ratings to
remain largely fixed and static. If, however, either the capital structure or
investment decisions are fixed and static, ratings must move, sometimes fairly
often, to reflect new risk. If both capital structure and investments are fixed,
ratings become the chief mechanism of reflecting risk to investors.
Since corporations can change their capital structures and investment
strategies while corporate debt is outstanding, it is fine to statically rate a firm
with a dynamic asset portfolio and dynamic capital structure. Static ratings are a
problem for mortgage pool debt, that is, RMBS, however, because mortgage
pools are static firmsthat is, static portfolios of investments funded by static
capital structures. Mortgage pool trustees cannot change their investment
strategies or alter their capital structures to ameliorate new or unforeseen risk
without changing the legal nature of the investment arrangements.
It is easy to see that the limitations of ratings for mortgage pools with static
capital structures and investment strategies have the most pernicious effects in
markets where a high degree of opacity renders ratings the chief mechanism of
monitoring and evaluating risk.
Furthermore, the pernicious effects of static capital structure and investment
strategies are heightened when the static firms invest in fixed income assets like
mortgages and other types of consumer loans. The problem arises because fixed
income asset performance is bounded from above either the investor gets what
they originally contracted for (repayment of principal and interest) or takes a
loss. Unlike real investments in, say, equity, since the firm has no opportunity for
substantial upside gain the firm has no opportunity to build additional earnings
buffer than can help offset unforeseen risk.
In light of these observations, then, the minutiae of the RMBS and CDO
securities merely obfuscate a real potential for inadequate risk transfer that can
arbitrage regulations anchored on credit ratings, like ERISA and Basel II.
52 Joshua Rosner & Joseph R. Mason
RMBS are sets of securities used to finance the purchase of a mortgage pool
from a mortgage originator. Like any firms securities, RMBS therefore pass the
risk of mortgage pool payments to individual investors.
Standard pass-through securities merely pass principal and interest payments
through to investors on a periodic pro rata basis. The creation of RMBS through
tranching, however, creates a much more complex set of securities. While it is
difficult to dispute that pass-through securities adequately pass mortgage pool
risk through to investors, it is much more difficult to estimate whether the
complex set of tranched RMBS securities passes the risk of the entire mortgage
pool through to investors. There are two primary reasons for that difficulty. First,
mortgage pools themselves are difficult to value due the commingling of default
and prepayment risk. Second, complex arrays of tranched RMBS are often thinly
traded, making it difficult to reconcile the total risk of the RMBS structure back
to the mortgage pool on the basis of market estimates.
Those two sources of the difficulty correspond to the two sides of the most
basic mathematical representation of portfolio risk. Standard MBA textbooks
present the simple two-asset portfolio standard deviation as:
n n n
p = wi2 i2 +
i =1 i =1
w w
j =1
i j i r
j ij
where the left-hand side of the equation, p, represents the total risk of the
mortgage pool. The right-hand side of the equation represents the total risk of the
portfolio of different tranched RMBS sold against the mortgage pool, where i
are the individual RMBS standard deviations and wi are the weights (tranche
sizes) of different RMBS constructed to finance the pool. The variable rij in the
equation is the correlation coefficient between the different classes of RMBS.131
One thing we know for sure: the total pool risk p does not go away by
merely being represented by a more complex set of securities. The sum of
portfolio risk in the securities sold to finance the mortgage pool therefore has to
equal the original risk of the mortgage pool. Thus, it must be that the right-hand
side is composed of securities that are specifically financially engineered to have
i and rij properties appropriately weighted wi in such a way as to accurately
reproduce the overall mortgage pool risk.
It is incredibly difficult to say, however, just what the right-hand side of the
above equation amounts to in real-world RMBS structures. If the actual structure
of RMBS securities is sold to investors on the basis of a combined portfolio
standard deviation that is less than p, risk has been masked. If the actual
structure of engineered securities and their weights is sold to investors on the
basis of a standard deviation more than p, risk has been overweighted (and the
deal will be uneconomical) to the issuer.
The problem of accounting for risk therefore becomes an exercise in
evaluating i and rij to make sure that the total risk of the set of engineered
132. Default risk is the risk that a borrower will not repay, on time and in full, all
principal and interest as promised in the financial instrument.
133. FitchIBCA, Subprime Collateral Trends and Early Payment Defaults, Apr.
13, 2007.
54 Joshua Rosner & Joseph R. Mason
Source: Charles Calomiris & Joseph Mason, Endogenous and Exogenous Mortgage
Prepayments in an Optimal Stopping Framework (Working Paper 2007).
134. Charles Calomiris & Joseph Mason, Endogenous and Exogenous Mortgage
Prepayments in an Optimal Stopping Framework (Working Paper 2007).
May 2007 Mortgage-Backed Security Ratings 55
mortgage accrued at the difference between the rate on the prepaid mortgage and
rates available at prepayment) amount to just over $576 million while interest
losses due to defaults amount to only about $12 million.135
The problem with prepayment risk is that, unlike default risk, there is no
industry standard to measuring prepayment risk. Public Security Association
(PSA) models attempt to correct for prepayment risk but do not estimate it
directly. Figure 22 illustrates common prepayment vectors from Fitch IBCAs,
Rating U.S. Residential Subprime Mortgage Securities.136 Note that ARM
prepayment vectors spike considerably at the end of the fixed period.
Source: Fitch IBCA, Rating U.S. Residential Subprime Mortgage Securities, Jul. 18, 2001.
Prepayment risk corrections affect the probable maturity of the bonds that
RMBS investors buy. When investors purchase RMBS securities, they are
therefore not only inferring default risk of mortgages in the pool, but also
inferring some probable maturity of their investments (when funds are repaid and
interest stops) as well.
Prepayment inferences about mortgage prepayments are based on ad hoc
corrections, not predictive prepayment models. The problem with formal
modeling is that unlike default risk, which is inversely related to economic
performance in a straightforward fashion, the double-edged sword relationship of
interest rate risk and prepayment risk creates a non-linear estimation
environment. Calomiris and Mason address the problem by estimating two
separate components of prepayment risk: one associated with prepayments
related to falling interest rates, and another associated with prepayments related
to other outside circumstances (like job transfers).
Calomiris and Mason call these two separate components of prepayment risk
endogenous and exogenous prepayments. Exogenous prepayments are
predicted as a function of attributes of the borrower and his or her local
135. None of this is meant to suggest that default risk is not as important as
prepayment risk. Total default losses (including losses on selling the home as collateral)
in the Calomiris and Mason sample amount to $4 billion. The main point, however, is
that repayment risk is costly and prevalent.
136. FitchIBCA, Rating U.S. Residential Subprime Mortgage Securities, Jul. 18,
2001.
56 Joshua Rosner & Joseph R. Mason
The main point of this section is to show that the most basic steps in valuing
mortgage pools, and thus RMBS, are inherently difficult. Some elements of
valuation, like measuring default risk, are thought to be relatively clear-cut and
utilize accepted industry-standard modeling procedures. Other elements of
valuation, like measuring prepayment risk, are not well-understood and therefore
require subjective adjustments to assumptions about mortgage pool behavior.
While there are existing industry standards for correcting for prepayment risk,
there are not the types of well-developed models for prepayment risk such as
there are for default risk. Hence, even the most basic elements of mortgage pool
risk, the left hand side of the risk equation at the beginning of the section, are
difficult to value.
RMBS are complex structured finance securities used to fund the mortgage
pools described above, with all the pools inherent default and prepayment
valuation difficulties. The complexity of structured finance products is meant to
smooth out some of the difficulties inherent in mortgage pool valuation to sell a
range of securities with different risk-return qualities to investors who value
those specific qualities (and tend to care less about other attendant valuation
difficulties). The inherent complexity of RMBS, coupled with fundamental
changes to underwriting and servicing standards, however, can also be used to
mask adverse changes to mortgage pool origination and performance and
therefore pose risk to RMBS investors and therefore risk to funding for socially
and economically important consumer mortgage originations.
The complexity of RMBS has increased as the statistical predictability of
collateral performance has become more established.
May 2007 Mortgage-Backed Security Ratings 57
Source: Maciej Firla-Cuchra & Tim Jenkinson, Why are Securitization Issues Tranched? (Oxford
University Working Paper 2005).
Figure 23 shows how RMBS and other structured finance products have
increased in complexity over time.137 The first private asset-backed security was
an RMBS issued by Bank of America in 1977, consisting of a simple pass-
through structurethat is, one tranche. As the industry matured and investors
became more comfortable predicting mortgage performance, RMBS became
substantially more complex.
Table 2 shows that other collateral types that began to be securitized well
after mortgages are far less complex. The first non-mortgage securitization was
equipment leases, followed by credit cards and auto loans, and more recently,
home equity, lease finance, manufactured housing, student loans, and synthetic
structures. All of those types of collateral illustrate tranching structures that are
measurably simpler than those for RMBS.
It is common to claim that RMBS became more complex as underwriting
technology became more efficient. Two examples of this are the movement from
standalone trusts to master trusts and the de-linking of tranches through issuance
trusts. Standalone trusts are simple and efficient for individual securitizations, but
woefully inefficient for repeated securitizations. The reason that standalone trusts
are inefficient for repeated securitizations is that each time an issuer wants to sell
more loans, it must create a new legal trust structure to house the loan pool and
issue the notes backed by that pool. Establishing an identical new trust structure
each time a new pool of loans is sold (often at least four times a year) repeatedly
incurs substantial fixed legal and administrative costs. Hence, a means by which
a firm could establish a shelf registration system for securitized assets
contributed a great deal of efficiency to the practice of securitization. By the late
1990s, therefore, most assets were being securitized in master trust structures.
Issuance trusts relieved the need to sell the entire structure of tranches at the
same time. In the past, the entire series of tranches had to be sold at once to
ensure that the lower-tier (junior) tranches provided the desired credit support for
the higher-tier (more senior) tranches. The problem is that widened credit spreads
resulting from adverse market shocks could (and did, during the 1998 Russian
bond default), occasionally shut down issuers ability to securitize assets at their
regular funding cycle. Those events led to legal developments that allowed
lower-tier tranches to be sold any time prior to higher-tier tranches as long as
higher-tier tranche maturities remained within those of the lower-tier securities.
Relieved of the constraint to sell all tranches of securities in the same market
environment, issuers could take advantage of favorable market conditions to
provide certainty to their regular funding cycle.
May 2007 Mortgage-Backed Security Ratings 59
Source: ABSnet.
RMBS are currently the most mature and complex of consumer structured
finance products. Figure 24 shows some representative RMBS structures. The
examples show that RMBS routinely contain many tranches of securities.
Furthermore, those tranches include even more sophisticated features to adjust
for prepayment and other inherent mortgage pool risks, including complex and
difficult-to-value securities like interest-only and principal-only strips, sequential
pay securities, and planned amortization class bonds.
C. The Complexity of Valuing the Pool and the Array of RMBS Backed by the
Pool Makes it Nearly Impossible to Trace the Accuracy of the Risk Transfer
in the RMBS Market
The main point is that the complex arrays of RMBS represented in Figure 24
need to be constructed to meet the constraints that the total pool risk be
represented in the entire (very complex) structure of RMBS. Earlier we
represented the mathematical representation of a portfolio of securities as:
n n n
p = wi2 i2 +
i =1 i =1
w w
j =1
i j i r
j ij
60 Joshua Rosner & Joseph R. Mason
Recall that the left-hand side of the equation, p, represents the total risk of
the mortgage pool. The right-hand side of the equation represents the total risk of
the portfolio of all the different tranched RMBS sold against the mortgage pool,
where i are the individual RMBS standard deviations and wi are the weights
(tranche sizes) of different RMBS constructed to finance the pool. The variable rij
in the equation is the correlation coefficient between the different classes of
RMBS.138
But this representation of risk is valid for only two classes of RMBS
securities, say equity and one class of debt. In practice, there can be up to 50 or
60 classes of debt and equity in a typical RMBS structure, each issued in
different proportions (weights) and with different, time-varying correlations (rij),
making it very difficult to estimate the true risk on the right-hand side.
Furthermore, we also showed that even the left-hand side is difficult to value,
given the complexities of default and prepayment risk. In that light, overly
complex arrays of RMBS can arguably add as much opacity to pool value as they
do efficiency to investors.
One thing we know for sure: the total pool risk p does not go away by
merely being represented by a more complex set of securities. The sum of
portfolio risk in the securities sold to finance the mortgage pool therefore has to
equal the original risk of the mortgage pool. Thus, it must be that the right-hand
side is composed of securities that are specifically financially engineered to have
i and rij properties appropriately weighted wi in such a way as to accurately
reproduce the overall mortgage pool risk. To do otherwise would be a
fundamental misrepresentation of risk.
In summary, RMBS are complex securities that are difficult to value.
Furthermore, RMBS are built on the backs of pools of mortgages, which
themselves are complex and difficult to value. Misrating RMBS and other
structured finance securities a substantial sector of the U.S. consumer credit
market at risk. Furthermore, fundamental changes to underwriting and servicing
standards are not easily identifiable in the inherent complexity of mortgages and
RMBS, posing risk to funding for socially and economically important consumer
mortgage originations. Below we demonstrate that the increased risk is magnified
by increasing concentration of CDO investments in lower-tranche RMBS
investments.
138. Recall further that if diversification is to be beneficial, r must be less than one,
so that the portfolio standard deviation will be below the mere sum of the standard
deviations of its component assets. Given that the value of all the RMBS is derived from
the same pool of mortgages, it is hard to argue that rij is substantially less than one in
many states of the world.
May 2007 Mortgage-Backed Security Ratings 61
Source: FitchIBCA, Rating Stability of Fitch-Rated Global Cash Mezzanine Structured Finance
CDOs with Exposure to U.S. Subprime RMBS, Apr. 2, 2007.
have the most pernicious effects in markets where a high degree of opacity
renders ratings the chief mechanism of monitoring and evaluating risk.
As before, the pernicious effects of static capital structure and dynamic
investment strategies are heightened when the static firms invest in fixed income
assets. The problem arises because fixed income asset performance is bounded
from above either the investor gets what they originally contracted for
(repayment of principal and interest) or takes a loss. Unlike real investment
projects, since the firm has no opportunity for substantial upside gain the firm has
no opportunity to build additional earnings buffer than can help offset unforeseen
risk.
Hence, CDOs, structured with a dynamic investment strategy but a fixed
capital structure, fixed income investments, and no market price signals require
dynamic ratings in order to adequately reflect risk. The main point is therefore
that, in the case of CDOs, bond ratings are again being used for something they
were not initially intended. In fact, ratings agencies continue to expand the
misuse of traditional ratings methods. The April 2007 IMF Global Financial
Stability Report warned that one of the most significant global risks outstanding
is the way that, the rating agencies continue to expand the application of their
ratings beyond the traditional credit risk domain.141
A. The Complexity of Valuing the Pool of RMBS and other Investments and the
Array of CDO Backed by the Pool Makes it Nearly Impossible to Trace the
Accuracy of the Risk Transfer in the CDO Market
Now the left-hand side of the equation, p, represents the total risk of the RMBS
and fixed-income pool. The right-hand side of the equation represents the total
risk of the portfolio of all the different tranched CDOs sold against the pool,
where i are the individual CDO standard deviations and wi are the weights
(tranche sizes) of different CDO securities constructed to finance the pool. The
variable rij in the equation is the correlation coefficient between the different
classes of CDO securities.142
CDOs again aggregate investments that, themselves, are difficult to value.
While CDOs differ in construction by type of collateral and purpose, most begin
with structures similar to RMBS. JP Morgans CDO Handbook (2001) illustrates
typical CDO tranche structure composed of some fixed and floating A and B
tranches supported by underlying C, D, and equity tranches. The top (A) tranches
are typically rated triple- or double-A, the next tier tranches (B) are rated single-
A, the next (C) triple-B, then double-B, with the equity typically not rated.
CDO funding structures, like RMBS, attempt to issue as many AAA-rated
securities as they can sell given the inherent risk in the underlying collateral.
CDOs attempt to issue as much class A, AAA-rated securities as possible
because those provide the deal with the cheapest funding. Lucas, Goodman and
Fabozzi suggest that typically about 77 percent of the securities in the structure
are class A securities and rated AAA, typically paying coupons about LIBOR
plus 26 bps.143 About 9 percent of the structure is class B securities, rated A and
paying about LIBOR plus 75 bps. The typical 2.75 percent of class C securities
are rated BBB and pay about LIBOR plus 180 bps. The riskiest rated bonds, the
class D securities, typically comprise about 2.75 percent of the structure, are
rated BB, and pay coupons of about LIBOR plus 475 bps. Equity, typically about
8 percent of the structure, is unrated and receives the residual cash flow from the
deal.
142. Recall further that if diversification is to be beneficial, r must be less than one,
so that the portfolio standard deviation will be below the mere sum of the standard
deviations of its component assets. Given that the value of all the RMBS are derived from
the same pool of mortgages, it is hard to argue that rij is substantially less than one in
many states of the world.
143. DOUGLAS L. LUCAS, LAURIE S. GOODMAN & FRANK J. FABOZZI,
COLLATERALIZED DEBT OBLIGATIONS: STRUCTURES AND ANALYSIS (2nd ed. Wiley
2006).
64 Joshua Rosner & Joseph R. Mason
Figure 26 shows actual CDO tranche structures. All the CDOs included in
Figure 26 are re-securitizations (also called structured finance, or SF, CDOs).
While most structures look similar to that RMBS, some, including those
illustrated in columns 3 and 8 are radically different.
The different funding structures reflect how CDOs fundamentally differ from
RMBS. In particular, CDOs are different from RMBS in at least six different
ways. First, whereas RMBS ABS are supported by static pools of underlying
assets, CDO pools are dynamically managed. Hence, the composition of the asset
portfolio can change dramatically through the duration of the CDO transaction.
Second, CDO transactions close before the pool of underlying assets is fully
formed. This aspect may be a benefit or a drawback. Beneficially, the manager
may be able to include in the pool greater diversity of collateral across industry,
credit, and vintage. As a drawback, however, like the classic corporate free cash-
flow problem, investors cannot be sure the manager will act as intended upon
investment. Third, CDOs are quite heterogeneous with respect to granularity.
Some CDOs may contain as little as twenty underlying assets, while others may
contain several hundred.144 Furthermore, several hundred underlying assets from
the same sector, that is, RMBS, does not add true diversification to the pool,
leading many in the industry to question the relevance of the traditional
calculations of diversity scores in contemporary CDOs.145 Because of the lack
of diversification, traditional actuarial loss methods applied to RMBS pools are
not properly applicable to CDO pools. Fourth, CDOs may illustrate more ratings
144. Note, however, that even several hundred underlying assets is still a relatively
small number compared to the hundreds of thousands of accounts underlying RMBS and
ABS pools.
145. See, e.g., Frank Partnoy, How and Why Credit Rating Agencies Are Not Like
Other Gatekeepers, in FINANCIAL GATEKEEPERS: CAN THEY PROTECT INVESTORS?,
(Yasuyuki Fuchita & Robert E. Litan, eds. Brookings 2006).
May 2007 Mortgage-Backed Security Ratings 65
146. See, e.g., S&P Global Cash Flow and Synthetic CDO Criteria, Mar. 21, 2002,
at 13-14.
147. See, e.g., S&P CDO Spotlight: Update to General Cash Flow Analytics
Criteria for CDO Securitizations, Oct. 16, 2006.
66 Joshua Rosner & Joseph R. Mason
B. Ratings Arbitrage: CDO Ratings Methods Are Looser than RMBS Ratings
Methods, Even when CDOs are Solely made up of RMBS
The advent of CDOs in the mid-1980s represented the first time ratings
agencies became consumers of their own ratings. With CDOs, ratings agencies
used published ratings on the constituent debt instruments in the collateral pool
along with the weights of different instruments in the capital structure, to rate the
CDO obligations produced on the basis of that pool. This was not too much
trouble in the beginning, because the investments carried in CDO pools were
almost exclusively standard corporate debt.
In the late 1990s, however, CDO pools began to carry a much wider variety
of debt from a much wider variety of sectors. Upon researching historical loss
incidence and severity across those different sectors, it was soon discovered that
the disparate sectors exhibited widely disparate performance.
A number of disclosure and arbitrage issues arose in attempting to properly
measure and adjust for the widely disparate performance. The disclosure
difficulties can be shown through the example of S&Ps experience (noting that
S&P is not unique in this regard). Nomura, in their Bond Rating Confusion,
report, writes, When the agency embraced a simulation-based rating
methodology for CDOs in 2001, it started using different implied asset default
rates for ABS and corporate bonds.149 However, the agency refrained from
disclosing a complete table of default rates over time for the separate asset
classes. Several months later, when S&P published its updated criteria for cash
flow and synthetic CDOs, it continued to disclose only a partial table.150 By not
148. Janet Tavakoli, GARP Risk Review Issue 22, January/February 2005 (If
everything else remained the same, but 2% of the portfolio defaulted, slightly more than
two percent of the first AAA tranche would not be deemed AAA. The AAA of the
second CDO presents a different picture, because 40% of the formerly AAA tranche
would no longer be deemed AAA.).
149. Sten Bergman, CDO Evaluator Applies Correlation and Monte Carlo
Simulation to the Art of Determining Portfolio Quality, S&P Special Report, NOMURA,
Nov. 12, 2006, at 6.
150. S&P, Global Cash Flow and Synthetic CDO Criteria, Mar. 21, 2002, at 46.
May 2007 Mortgage-Backed Security Ratings 67
releasing complete tables of default probabilities for ABS and corporate bonds,
S&P temporarily sidestepped the problem of having multiple definitions
associated with its rating symbols. Later, when S&P addressed the treatment of
municipal bonds in CDOs, it acknowledged the stronger historical performance
of the municipal sector but it still refrained from publishing tables with different
idealized default rates for the different sectors.151
Figure 27 shows the default probabilities from version 3.2 of S&Ps software
for time horizons of three, five, and seven years. Figure 27 shows that over time
151. S&P, Finance Criteria Book, Apr. 12, 2005, at 305-307; Nomura, Bond
Rating Confusion, June 29, 2006, at 5-6.
68 Joshua Rosner & Joseph R. Mason
horizons of both five years and seven years, S&P ascribes a higher default
probability to a CDO rated AA than to an ABS rated A. Over a three year time
horizon, a CDO rated AA has a higher probability of default than an ABS rated
A-. The tables, therefore, have some bizarre implications. According to Nomura,
Suppose you have a seven-year ABS rated AA+. According to the tables, the
instrument has an idealized default probability of 0.168%. If we repackage the
security (all by itself) and call the repackaged instrument a CDO, it ought to get a
rating of AAA because the idealized default rate for the AAA-rated CDOs is
0.285% over seven years.152
The ability to repackage financial securities and call them something else,
with no fundamental change to their risk characteristics, in order to achieve an
improved bond rating is the fundamental source of ratings arbitrage. As long as
ratings agencies mean different things when referring to CDOs, ABS, and
Corporate debt, incentives will continue to be skewed by risk arbitrage.
Furthermore, embedding ratings into regulation through ERISA and Basel II only
worsens the incentives to use opacity to the issuers benefit (and the investors
loss).
The problem with the complex arrays of securities originated as CDOs is that
the CDO sector has grown so immensely in a short period of time. Figure 28
illustrates that annual issuance of CDOs has grown from nearly zero in 1995 to
over $500 billion in 2006. In fact, CDO issuance is growing so fast that new
issuance in 2006 amounted to approximately the total of the three preceding
years summed together.
450
400
350
300
250
200
150
100
50
0
1987 1989 1991 1993 1995 1997 1999 2001 2003 2005
Figure 28 also shows the risk of relying on CDO markets for funding growth
of underlying collateral underwriting. Figure 28 shows that CDO issuance
dropped off precipitously from 1998 to 2002, exceeding 1998 levels again only
in 2003. That drop-off arose from a combination of economic conditions along
with unforeseen difficulties in the corporate loan and bond markets and
manufactured housing, aircraft lease, franchise business loan, and 12-b1 mutual
fund fee structured finance sectors that accounted for a significant component of
CDO investments at the time. CDOs have since moved out of those sectors, and
into RMBS and commercial mortgage-backed securities (CMBS).
The kind of market dynamics illustrated in 1998 suggest the potential for
high volatility in sectors that rely principally upon securitization and re-
securitization through CDOs for funding.
The potential for high funding volatility arises because of the structure of
RMBS and the types of RMBS that CDOs hold. The structure of RMBS
providing funding for a specific pool of mortgages is composed of an array of
debt securities that pay investors principal and interest sequentially. If enough
principal and interest remains after paying the first class of investors, the second
is paid. If money remains after that, the third class is paid, and so on, and so
forth. In this manner, the last class of securities to get paid bears the majority of
the risk. Without investors willing to purchase the last classes of securities
typically, about 10 percent of the securities sold the prior classes bear the risk.
If the prior classes are not willing to bear the risk, the other 90 percent of the
70 Joshua Rosner & Joseph R. Mason
mortgage pool cannot be funded, that is, the mortgage originator cannot sell the
loans. Hence, the last classes of securities are providing about 10:1 leverage for
the structure of RMBS, so that every dollar of lower-tranche RMBS supports
about $10 of higher-tranche RMBS.
Specific types of CDOs have arisen that specialize in holding such risky
classes of securities have become popular in recent years, providing a great deal
of funding for lower-tranche RMBS at that 10:1 leverage ratio. But since so
much lower-tranche RMBS has been sold to those CDOs, the CDO sector now
holds a very high concentration of mortgage risk. If RMBS begins to default and
the CDO sector restricts investments in RMBS or shifts out of RMBS, the
mortgage industry loses funding of $10 for every $1 CDOs no longer want to
buy. This scissors effect of leverage produces very volatile effects on the
market that is leveraged, in this case, the mortgage market.
So just how much lower-tranche RMBS do CDOs hold? The FDIC reports
that 81 percent of the $249 billion of CDO collateral pools issued in 2005, or
$200 billion, was made up of residential mortgage products. (FDIC Outlook, Fall
2006) Moodys CDO Asset Exposure Report for October 2006 reveals that 39.52
percent of the collateral within the 678 deals covered by Moodys consists of
RMBS, just over 70 percent of that in subprime and home equity loans and the
other 30 percent in prime first-lien loans. Figure 29 confirms similar magnitudes
from a smaller sample of CDOs monitored by Fitch.
Source: FitchIBCA, Rating Stability of Fitch-Rated Global Cash Mezzanine Structured Finance
CDOs with Exposure to U.S. Subprime RMBS, Apr. 2, 2007.
It is clear that CDOs hold a lot of RMBS, but what matters is how much of that is
lower-tranche RMBS, which is the source of the scissors effect. Moodys CDO
Asset Exposure Report for October 2006 reveals that 70 percent of collateral in
the pools underlying the 2005 resecuritization CDO vintage was below AAA-
rated, and the largest ratings cohort, at 40 percent, was Baa. About 10 percent of
the 2005 vintage collateral pool was rated below Baa. Overall, about 75 percent
of collateral in the pools underlying resecuritization CDOs was below AAA-
rated, the largest cohort being, again, Baa at 42 percent. About 16 percent of
collateral was rated below Baa. Table 3, from Moodys, shows that overall
May 2007 Mortgage-Backed Security Ratings 71
weighted average rating factor (WARF) for the top collateral types for the
resecuritizations is 225, which equates to about a Moodys Baa1 rating.
Source: FitchIBCA, Rating Stability of Fitch-Rated Global Cash Mezzanine Structured Finance
CDOs with Exposure to U.S. Subprime RMBS, Apr. 2, 2007.
FIGURE 31: HOW MUCH AND WHAT KIND OF RMBS ARE IN CDOS?
How large of an effect? SIFMA estimates that about $1,326 billion was
issued in private RMBS in 2005. If 10 percent of that is lower-tier (junior)
securities, then about $133 billion in lower-tier securities supported the rest of
the $1,193 billion.
Hence, the CDO market purchased more mezzanine RMBS in 2005 as was
actually issued that year. Furthermore, and crucially, the relatively small amount
of RMBS purchased by the CDO market provided support for the rest of the
$1,193 billion issued in private RMBS during the entire year of 2005.
In summary, the CDO market adds liquidity to the RMBS market in a highly
leveraged fashion by funding lower-tranche RMBS securities, and the experience
of the ABS markets in the early 2000s illustrates that the liquidity provided by
CDOs is very fragile. As is common in highly leveraged markets, the scissors
effect of leverage can create substantial market volatility when investors enter
and leave a highly leveraged market quickly. Since CDO difficulties arising from
RMBS ratings deterioration is expected to be the proximate cause of mortgage
market de-leveraging, we turn to analyzing the relationship between RMBS
ratings and CDO ratings.
The main problem we have described is that the RMBS market is built upon
the shaky statistical predictability of mortgage pool performance. The CDO
market then builds upon the shakier foundation of the statistical predictability of
RMBS performance to provide additional market liquidity.153
Preliminary evidence for this view lies in the relationship between ratings
changes in CDO markets and underlying collateral, that is ABS and RMBS,
markets. Given recent events, we now know that there are substantial numbers of
defaults and near-defaults in RMBS mortgage pools. It is only a matter of time,
therefore, before they accumulate to levels that will threaten rated mezzanine
RMBS that is the investment of choice in many CDO pools. Many people are
therefore asking when the RMBS defaults will hit CDO returns. In an opaque
market prices may already be at distress levels we wont know, however, until
investors try to sell.
We can infer when the defaults will hit CDO performance by reviewing the
lessons already learned from the NERA report in 2003. The NERA report
analyzed the effects of notching in the ratings industry. Notching refers to the
industry practice whereby one agency adjusts ratings of structured finance
collateral from other agencies for the stated reasons of (1) bringing them in line
with ratings it believes it would have assigned to the collateral and (2) adjusting
for uncertainty and perceived differences in monitoring practices.154 The study
came about at a time when ratings agencies were increasingly asked to rate
collateralized debt obligations with underlying collateral pools that include
structured finance securities rated only by other rating agencies.
The NERA report is important for understanding present conditions in CDO
markets because it points out the need for one ratings agency to do substantial
additional research before integrating ratings of another agency that relate to, say,
underlying collateral in a CDO. First, the outside rating change must be adjusted
to be comparable with the CDOs ratings agency. Then, the CDOs ratings
agency needs to analyze the CDOs cash flow implications of the change in the
single underlying collateral instrument. Last, the CDOs ratings agency needs to
decide whether to take action on the CDO itself. All those steps take additional
time when the structures of the securities of concern are more complex.
In footnotes to their CDO Asset Exposure Report, Moodys notes that it
can take anywhere from three to seven weeks to normally incorporate another
ratings agencys change into their own CDO ratings.155 Hence, it would be
expected that CDO ratings changes considerably lag RMBS and ABS ratings
changes due to opacity between markets and across ratings agencies.
The CDO market is opportunistic in the way it drops collateral types that
are out of favor with investors and picks up collateral types that are in
favor with investors. The best example of this is the switch out of poor-
performing high-yield bonds and into well-performing high yield loans
between 2001 and 2003. Also, certain types of ABS present in SF CDOs
from 1999 through 2001 disappeared from later vintages: manufactured
housing loans, aircraft leases, franchise business loans, and 12b-1 mutual
fund fees. All of these assets had horrible performance in older SF CDOs.
In their place, SF CDOs have recently focused more on RMBS and
CMBS.157
Hence the degree of leverage inherent in CDO funding, along with the potential
for high volatility in that funding, introduces the potential for public policy
issues.
E. Domestic Links between Mortgage Markets, RMBS, and CDO Funding and
the Housing Sector and Economic Growth
FIGURE 34: CDO ISSUANCE AND U.S. HOME EQUITY ABS INDICES
161. Jeffery DAmato, Risk Aversion and Risk Premia in the CDS Market, BIS Q.
REV., Dec. 2005.
80 Joshua Rosner & Joseph R. Mason
Source: Jeffery DAmato, Risk Aversion and Risk Premia in the CDS Market, BIS Q. REV., Dec.
2005.
So far we have seen many of these expectations play out in markets already.
Ratings agencies continue to say that the worst is over, only to revise their
assumptions downward over and over. By late March, S&P was raising their loss
assumptions on 2006 vintage subprime loans another 125 basis points, to 7.75
percent overall. At the same time, S&P continued to maintain in reports that most
investment-grade bonds originated last year would remain protected, even as
underlying loan performance would be the worst in history. Most can only be
May 2007 Mortgage-Backed Security Ratings 81
162. S&P Raises 2006 Subprime Mortgage Loss Expectation, REUTERS, Mar. 27,
2007.
163. Serena Ng, Subprime Cloud Overshadows S&P, Moody's, WALL ST. J., APR.
24, 2007, AT C1.
164. Id.
165. Vikas Bajaj, Defaults Rise in Next Level of Mortgages, N.Y. TIMES, Apr. 10,
2007.
166. Jody Shenn, Commercial-Mortgage Risk Forcing Change, Moodys Says,
BLOOMBERG, Apr. 11, 2007.
167. Buddy, Just Hand Over That Dime, ECONOMIST, Mar. 31, 2007; Standard &
Poors Warns of Loose Standards in Commercial Lending: Flood of Liquidity into
Interest-only Loans Cited, INMAN NEWS, Apr. 11, 2007.
82 Joshua Rosner & Joseph R. Mason
Source: IMF Global Financial Stability Report Statistical Appendix, Apr. 2007, Ch. 1, pp. 34-5.
168. Brian Blackstone, Greenspan Sees Spending Link to Home Equity, WALL ST.
J., Apr. 24, 2007; Abby Goodnough, Housing Slump Pinches States in Pocketbook, N.Y.
TIMES, Apr. 8, 2007.
169. IMF Global Financial Stability Report Statistical Appendix, Apr. 2007, at 36.
170. Christopher Conkey and Phil Izzo, Consumers May be Getting Tired: Weight
of Gasoline Prices, Housing and Job Markets is Piling Up, WALL ST. J., May 9, 2007.
May 2007 Mortgage-Backed Security Ratings 83
Source: IMF Global Financial Stability Report Statistical Appendix, April 2007, Ch. 1, p. 142.
171. Peter Garnham, Dollar Continues to Lose Ground, FIN. TIMES, Mar. 15, 2007.
84 Joshua Rosner & Joseph R. Mason
that will result from lower liquidity in the RMBS sector will further weaken
credit spreads and depress CDO and RMBS issuance. This feedback mechanism
can create imbalances in the U.S. economy that, if left unchecked, could lead to
prolonged domestic economic implications for U.S. standing in the world
economic order.
The potential for prolonged economic difficulties that also interfere with
home ownership in the United States raises significant public policy concerns.
Already we are witnessing restructurings and layoffs at top financial institutions.
More importantly, however, is the need to provide stable funding sources for
economic growth. The biggest obstacle that we have identified is lack of
transparency. The structural changes noted in our previous draft largely went
unnoticed by RMBS investors until only recently. We argue that those changes
went unnoticed largely because of the existing complexity and valuation
difficulties underlying todays RMBS markets.
But policymakers and ratings agencies are still reluctant to examine some of
the key frictions that have caused the present mortgage mess.
Congress is calling for increased loan mitigation without thought to the vast
heterogeneity in mitigation standards existing in todays market nor a sound
analysis of best practices in the field.
While the mortgage industry is pulling back from high LTV products, which
makes perfect sense in an economic climate with little home price appreciation,
legislators are pushing to expand the high-LTV sector, even with little hope of
owners building equity to incentivize repayment, much less any limit to limit
cash-out refinancing that arbitrage the mortgage interest tax deduction and erode
the buildup of equity on government-insured mortgage products.
Some are even proposing bailouts for overextended borrowers, ignoring the
approximately half of U.S. homeowners who own their homes outright through
their own sweat, perspicacity, and prudence.
And there is still no focus on monitoring bank funding markets. The feared
outcome is nothing less than a 21st century bank run, this time from CDO
investors rather than depositors. High yields in RMBS in the past several years
led to a massive infusion of CDO hot money into the RMBS sector in an
environment similar to that of the thrift crisis of the late 1980s. Like the thrift
crisis and its aftermath, therefore, recent events not only threaten these
institutions, but also threaten the U.S. consumer and taxpayer as well. Indeed,
current estimates of losses due to the mortgage mess are on par with the thrift
crisis.
Perhaps of greater concern is the reputational risk posed to the U.S. capital
marketsmarkets that have historically been viewed as among the most
transparent, efficient, and well regulated in the world. The economic value of
mortgage securitization and the risk transfer value of CDO issuance support their
further use. However, there should be significant resources allocated to building
the regulatory framework surrounding their structuring, issuance, ratings, sales,
and valuation. We believe that efforts to provide transparency to these new
product areas can foster stability while maintaining liquidity to the underlying
collateral sectors and supporting further meaningful financial innovation and
capital deepening.
May 2007 Mortgage-Backed Security Ratings 85
BIBLIOGRAPHY
Ambrose, B.W. and Capone, C. A. The Hazard Rates of First and Second
Defaults. Journal of Real Estate Finance and Economics, 20(3):275, 2000.
Brady, Peter J., Glenn B. Canner, and Dean M. Maki. The Effects of Recent
Mortgage Refinancing. Federal Reserve Bulletin 86:44150, 2000.
D Amato, Jeffery. Risk Aversion and Risk Premia in the CDS Market. BIS
Quarterly Review, December 2005.
Deng, Y a and Quigley, J and Van Order, R. Mortgage default and low
downpayment loans: The cost of public subsidy. Regional Science and Urban
Economics 26:263-285, 1996.
Harvard University Center for Housing Studies, State of the Nations Housing
2006, Home Page, available at http://www.jchs.harvard.edu/publications
/markets/son2006 /son2006.pdf (last visited February 2007).
King, M.J. Tougher Rating Criteria May Impact Issuers, Secondary Market
Executive, Vol2, Issue 1, 2006.
Lucas, Douglas L., Goodman, Laurie S., and Fabozzi, Frank J. Collateralized
Debt Obligations: Structures and Analysis, 2nd Ed. Hoboken, NJ: Wiley, 2006.
S&P CDO Spotlight: Update to General Cash Flow Analytics Criteria for
CDO Securitizations, October 16, 2006.
S&P Global Cash Flow and Synthetic CDO Criteria, March 21, 2002.